Workshop Day 1: Ultrasound Neuromodulation for Mental Health Applications
HOLLY LISANBY: Welcome to our NIMH workshop on ultrasound neuromodulation for mental health. It is my pleasure to welcome you and open this session. My name is Holly Lisanby and I direct the Division of Translational Research at NIMH.
We are especially excited about novel tools such as focused ultrasound neuromodulation that can help us study relationships between brain and behavior, ultimately understand the etiology of psychiatric disorders, and hopefully lead to novel interventions.
In particular, the ability of focused ultrasound neuromodulation to impact deeper brain areas without impacting superficial cortex is something that makes it unique among other tools and may have special advantages. There are also some particular research challenges and opportunities that we look forward to taking a deeper dive into throughout this workshop.
I would like to thank our workshop cochairs, Dr. Kimm Butts Pauly and Lennart Verhagen and also our NIMH organizing committee, Drs. Jessica Tilghman, Dr. Lizzy Ankudowich and Dr. Craig Fisher. And now without further ado I would like to introduce Jessica Tilghman, who will kick us off today.
JESSICA TILGHMAN: Good afternoon, everyone. On behalf of the NIMH organizing committee, Dr. Lizzy Ankudowich, Craig Fisher and myself Jessica Tilghman as well as our meeting chairs Drs. Kimm Butts Pauly and Leonard Verhagen, we thank you for joining the NIMH workshop Ultrasound Neuromodulation for Mental Health Applications.
Here I’m going to go over the workshop program for the next two days. There will be three sessions each today. Today’s first session will be a brief showcase of some of the latest research in the field. This session will be followed by a discussion on biophysical considerations. And we will end the day with a discussion on physiological and clinical considerations.
And then tomorrow we will continue with sessions on the regulatory pathway for focused ultrasound neuromodulation as a treatment, experimental design and planning, and optimizing target engagement: parameter space and effects. We will then conclude with a synthesis of the discussion held over the next two days, as well as opportunities and next steps for ultrasound neuromodulation research and mental health.
So the overall goal of this workshop is to share the latest findings and discuss best practices for research in focused ultrasound neuromodulation for neuropsychiatric disorders.
And some of the questions that we are aiming to discuss further include what is the current understanding of how ultrasound interacts with the brain, what are the risks, potential side effects and patient burdens, and how can they be minimized. Given the above, how do we define the therapeutic window to maximize effects. And how can we best plan our stimulation and design our trials to ensure rigor.
And other discussions will revolve around some of the main challenges facing the field, including the effect of parameter settings on biophysical and neurophysiological outcomes, how to quantify dose in a way that is relevant to neuromodulation, effects of repetitive protocols, accommodation of individual skull variability, off-target or confounding effects from activation of intermediate brain regions, and the impact of brain-state on neuromodulatory effects.
Lastly, as a reminder, the views and the opinions expressed in the following presentations do not necessarily represent the views of the National Institutes of Health, the Department of Health and Human Services, or the United States Government. And with that I would like to thank you all again for your attention, and I will turn it over to Dr. Grace Hwang to introduce our first session. Prior to joining NIH NINDS as a Program Director in the Brain Initiative, Grace was a BRAIN awardee who investigated in non-invasive ultrasound neurostimulation for recovery of plasticity in adult mice. Grace, take it away. Thank you.
Session 1. Showcase.
GRACE HWANG: Thank you Jessica for the introduction. And welcome everyone to the first showcase session. This session will feature hot topics in ultrasound stimulation in preclinical models and in ultrasound induced neuroplasticity in humans. Both presenters will describe their research in the context of the state of the art. Our first presenter is Dr. Keith Murphy from Attune Neuroscience. He will discuss ways to achieve remote control of brain activity in rodents.
Our second presenter is Dr. Elsa Fouragnan, who is joining us from the University of Plymouth in the United Kingdom. She will present about ultrasound induced early phase neuroplasticity in humans and implications for clinical translation.
So without further ado, Keith, the floor is yours.
Remote Control of Brain Activity in Rodents
KEITH MURPHY: Thanks so much, Grace. I appreciate the introduction. I would like to start with financial disclosures. I was an inventor on a patent application assigned to Stanford from which I’ve received royalties. And I’m also a shareholder and salaried employee of Attune Neurosciences.
And I thought I would begin this talk by saying what we’re all trying to achieve with preclinical models, and sort of the direction of how we’re building tools and what we’re thinking about in research.
And one of the main principles behind preclinical models is how can we obtain as much information from the brain as possible while delivering focused ultrasound. We also want to improve ultrasound stimulation precision. We want to know exactly where the focus is, and exactly what the size is.
And the last bit is, while we’re stimulating the brain, it is ideal that we capture the most naturalistic behavior as possible, and pathological states that really represent the disorders that we’re trying to design treatments for.
And I wanted to highlight a few other tools before I talk about my own work. Tala Lemaire and Shy Shoham have recently developed a small cortical window into the mouse brain integrated with a focused ultrasound tool, so that they can visualize individual neurons in the brain and how they respond to ultrasound.
Similarly, Hector Estrada and Daniel Razansky developed a multi-element array where they could steer around the entire cortical area of a mouse while visualizing the entire cortex.
Tamasa Dilanni and Brog Iran(ph.) have developed a high element count array, similar to the last tool, except this one allows the rats that are mounted with it to freely behave. So they can essentially steer around all these different brain regions while watching the animal’s behavior.
Elisa Konofague is taking a sort of different approach. She is working on tools to image the entirety of the deep brain cortex simultaneously with ultrasound, and she has integrated this with high intensity probes that also stimulate the brain at the same time. And on the righthand side the back draw image is changes in blood flow she’s measuring in ultrasound, with an overlay of where displacement is occurring in the brain.
Work by Rachel Niu and Ben He has been exploring plasticity in animal models. So here they stimulate a hippocampal circuit electrically while stimulating with focused ultrasound. And on the top right you can see that changes in the circuit’s evoked potentials last long after the stimulation is on. And Grace has been doing work in this field as well.
We’re not just advancing small animal tools. Taylor Webb and Yan Kubanek have also been developing a steerable array that they call Remus where they can target individual areas of the macaque brain, and they can monitor the exact position of the focus using temperature and MRI readouts.
Charles Casky’s group is also working on different methods for detecting focal position using a method called radiation force imaging that will be talked about later in the conference. And at the same time they are looking at functional blood flow in the macaque brain. So they can start to correlate the position of the focus and changes in network activity.
So in this sort of field of developments, we were looking to start to look at deep brain locations that were involved in sleep/wake behavior. And they are very specific cell types in the brain. Because we were looking originally at sleep and wake, we needed the animals to be able to behave very freely.
So we decided to turn to a tool called photometry, where you virally encode a light sensitive molecule into specific cell types in the deep brain, which is shown on the left, and you install a small optical fiber above those cells so that you can monitor when they’re active or when they’re suppressed.
And our very simple innovation was to create a ring transducer that could be easily sort of mounted onto the skull of the animal. And this ring would emit radiating waves that had a natural focus right at the cells which sit below the fiber. And this allows you to simultaneously image neural activity while animals are behaving.
We called the development fiber photometry coupled ultrasound or PHOCUS, and it’s a simple combination of piezoelectric components, matching layers, and mounting brackets.
So give you guys an idea of how animals can behave, here is an animal not being stimulated at the moment but freely walking around its cage with the device on. And you can see he can sort of reach up on the cage and wander around. You might think about how a lightweight device might be used in your own research, particularly in mental health applications that require complex behaviors.
We have recently tried to push the boundaries of what could be done with this tool, and we began studying rotarod activity. So here the mouse is sort of tasked with walking and balancing on this spinning wheel, and we’re able to show that you can use ultrasound to intervene in motor behavior.
An earlier subproject that we did included recording from the ventral tegmental area. That’s a deep region of the brain involved in stress, anxiety, and sleep. And we simultaneously recorded inhibitory GABAergic neurons in this area, and their collocal dopaminergic outputs. And what we found was that when we stimulated with FUS, which is the purple shaded area here, we got strong excitation of GABAergic neurons, while dopamine neurons seem to be suppressed, which sort of suggests that these two cells might be interacting locally. Because right after the FUS goes off and the GABAergic activity drops, you start to see that rise in dopaminergic activity, which is the green trace.
Another project we worked on focused on the hippocampal area, which is obviously involved in many different things, locomotion, spatial awareness, and more recently feeding behavior has been shown. And we simultaneously labeled two different cell types. So we looked at PV, perivalvar positive interneurons with the red fluorophore and CaMKII positive excitatory neurons with GCaMP6.
And although most protocols excited both cell types, we found at 900 hurts at a 20 percent duty cycle over five seconds was able to potently decrease activity of our excitatory neurons, while strongly increasing activity in the inhibitory cell types. Again, suggesting the possible interaction between these two in the area.
And because excitatory neurons represent such a large portion of this area, and were inhibited so strongly, we thought we might be able to visualize that in the context of the entire brain. So to do this we turned to PET imaging. And what we did was we injected a small animal with a radio labeled fleuro deoxy glucose and allowed it to walk around a cage for 30 minutes while stimulating every three minutes. And this was done at NIDA in Bethesda. After stimulation we put the animal into a small animal PET scanner, and we essentially visualized glucose uptake in the entire brain, and we compared the stimulated hippocampal area with the contralateral side of the brain.
And when you look at highly significant suppression, you can see that the area filled was exactly at the focus, as we had intended. Which was a little surprising, just how focal it was considering the size. But if you lower the significance, you start to get a less modulated circuit view, you can actually see the hippocampal tracks outside of the focus were also affected, which the red arrow is pointing to in the bottom right. So this suggests that our reduced glucose uptake was not just some mechanical effect at the focus, but also a synaptic step away, and that we could intervene in whole circuits.
It has previously been shown that stimulating inhibitory cell types in the hippocampus could suppress seizures. So we tried to examine this ourselves by damaging the hippocampus with kainite acid, which typically causes epileptiforms while stimulating the contralateral side of the brain with ultrasound. And on the righthand side you can see that we get nice epileptiform spikes, these high frequency oscillations.
And very typically when we would turn on the ultrasound, represented by the purple box, these seizures would halt almost immediately. And this effect would last somewhere between 10 and 30 seconds. So while it didn’t last too long, it suggests that we might be able to create closed loop interventions, where if we can detect a seizure starting to grow or come on we might be able to intervene in the brain in real time.
More recent work from the lab has reached into other areas of the brain. And we expansively searched many parameters that I won’t have time to talk about today, but I wanted to focus on one particular region, the dorsal medial hypothalamus. And when we manage GABAergic neurons in this region we found that they were extremely responsive to focused ultrasound pressure. On the bottom right you can see a ten second stimulation at a low duty cycle, and across different pressures. And you can see that not only do we get sustained activity during the entirety of the FUS delivery, but that it lasts for about a minute or so afterwards.
And when we monitored animals during the stimulation, so on the top is an infrared tracking of a mouse being stimulated, and the bottom is a photometry trace, you can see when the FUS comes on the animal immediately begins to start walking around its environment. And this lasts for a little bit after the focused ultrasound is turned off. Now, you might be thinking this is a really good example that Keith is showing here, but I can tell you that this behavior is extremely robust.
And here I’m showing the tracking from all the animals overlayed, and on the righthand side I’m showing a control region. So just one and a half millimeters away we’re moving the focus. And when the black box comes on, that’s the ultrasound stimulation. You can see robust walking from almost all the animals that reaches the time period after, showing that we’re able to remotely control walking behavior with just a short burst of ultrasound stimulation. And we’ve further validated that even regions 0.3 millimeters away don’t quite have this effect. So we think it’s very focally specific.
In reaching into other regions, so here we were looking at the center median nucleus of the thalamus, we found another striking bit of information. And this was that if you tweak parameters enough you can have very similar intensities and total energy delivered while achieving very different directionality in cell type response.
So the red trace shows a quote-unquote sort of excitatory protocol, where we have very high duty cycle, short duration, and very low pulse repetition rate, so very slow pulsing for a short period, we’re packaging energy very tightly. And the purple trace is a sort of quote-unquote inhibitory protocol where we’re doing just the opposite. So with the same amount of energy we can either excite or inhibit a brain region.
And when we looked at behavior in this brain region, we didn’t really see any changes in walking or anxiety, but we did notice head motion. So this indicates that the animal is sort of looking around the arena and aware of its environment. And in the bottom left you can see with the excitatory protocol we were able to increase this sort of alertness, while the inhibitory protocol could decrease it for most of the duration of the ultrasound delivery.
So why is it that these two protocols with similar energy would have different responses? We thought that temperature from packing energy over a shorter period might explain this effect. So we took our device, and where we normally have the optical fiber we replaced it with a thermocouple probe, and we spent a lot of time verifying that this was working properly, because when we looked at the skull temperature right below the transducer there was a gentle heating which you would expect, because the skull absorbs a lot of ultrasound. But at the focus we saw exactly the opposite.
So with our strongly exciting protocol we saw a drop in temperature by almost one degree, which is really substantial in the brain, and a much lesser effect with our inhibitory protocol. We thought that vasodilation might explain this effect, because blood can act sort of as a perfusion pump of temperature in the brain. So we injected rhodamine B dextran into the blood while simultaneously monitoring activity from these neurons.
And you can see that while we get this robust increase in neural activity shown in green, and a short spike in blood volume, which is probably something like the FMRI BOLD response, we get this lasting slow decrease in blood volume that follows. And we think what is happening is that we are getting modest vasoconstriction after this large excitation of neural activity. And this is not new, this is a phenomenon that has been explored in the past.
And it’s thought that the cyclo-oxygenase system detects large amount of neural activity in the brain and can vasoconstrict. So to test this we injected animals with either a vehicle or ibuprofen, a cyclo-oxygenase inhibitor. And while we didn’t see changes in neural activity, which is on the left, we almost completely eliminated this vasoconstrictive effect. You can see in grey the ibuprofen treated animals.
So we think that we’ve developed a protocol that is just generating a lot of activity and a lot of physiological effects, and we look forward to examining this further. And with that I’d like to thank my postdoctoral mentor at Stanford, Luis de Lecea, and all of our collaborators on this work, and our various levels of support from the NIH for these projects. Thank you.
Inducing Early Phase Neuroplasticity in Humans
ELSA FOURAGNAN: Thank you very much for the introduction, Grace. So today I’ll be focusing on humans and how TUS can be used to achieve short and even long-term changes by inducing neuroplasticity, which will be particularly important when we’re thinking about clinical applications.
So treating the brain with TUS can elicit diverse response, including acute effect, which will be happening during or immediately after a sonication, or delayed effect, where the outcome can be observed in different forms, in the form of local effects, including changes in certain concentrations of metabolite in the brain, but also remote effects, and functional effects, which could be in the form of behavioral or cognitive change in the hour that follows the intervention.
I will also be discussing today very briefly the potential for TUS to trigger longer term outcomes, which could be particularly relevant for clinical applications. So here I am thinking about changes that could be observed in the weeks or even months following intervention. But for the purpose of this talk I will mainly talk about early neuroplasticity that can be observed in the hour that follows the intervention.
And I will present evidence of chemical change in the sonicated region. I will also be talking about change in regional interaction using resting state network connectivity. And I will also be presenting some task-related behavioral change with an indicating of specific brain behavioral relationship, particularly relevant for learning and decision making.
So in the first study in the lab we were interested in the in vivo concentrations of GABA, which is the main inhibitory neurotransmitter in the brain, in the hour that followed the intervention. This is a pre-register study. And in this study we used a five hertz pattern TUS protocol before people were put into an MR scanner, and after that we just collected resting state spectroscopy data. And this was in the hour that followed the intervention.
We had 24 participants coming. They came four times. So this was within a repeated design protocol. They first came for a series of anatomical pictures that were used for creating acoustic simulation and temperature simulation to plan the intervention, and also for neural navigation on the day. And then they were asked to come three more times on different weeks, where they either received TUS on the dorsal interior cingulate cortex or on the posterior cingulate cortex, or sham. So on those three weeks they were either receiving a real sonication or sham.
And this is just the results of the simulation, where you can see the overlap between the spectroscopy voxel, the spectroscopy voxel is going to be in the vicinity of the focus region for both regions. And on the right image here I’m just presenting a representative subject where the focal volume, which will be full width half maximum is presented within the three different MRI voxels. So it’s hard to see, but there are actually three voxels that corresponded to the three different weeks where we acquired the data.
So in terms of data quality, both fit error and signal to noise ratio were very good for both regions. So we moved on with looking at the main analysis. And here are the results, which I would like to unpack a little bit for you. So here we’re looking at the PCC voxel, so the spectroscopy voxel in the PCC region. And this is the concentrations of GABA relative to water. So in the middle here is the condition where this brain region had received sonication.
On the right is when no sonication had been applied, and on the left after the ACC which is the control region for that one was stimulated. And you can see a very consistent decrease in GABA concentration following. So this was about 30 to 40 minutes after the sonication. So you can see a clear decrease in GABA. But this was not the case for the dorsal interior cingulate cortex, which is quite surprising given the fact that it is the same protocol and it is the same individuals.
In the resting state connectivity we observed, a change in the way the ACC covary with the rest of the brain after test. But this was less pronounced than what we also observed for the posterior cingulate cortex. So we can see different networks after 15- and 45-minutes post TOS, which were differentially coactivated with the region that was sonicated. Very similarly, when we looked at the contribution of the ACC and the PCC to their respective network so the PCC contribution to the default network was much more enhanced after sonication. And this was also the case after the stimulation of the DACC but to a lesser extent.
So this could be intriguing, what can explain these differences. Perhaps it has to do with the state of the brain. So you might know that when people are resting the default mode network is the principal network at play, and the PCC is very active in that network.
Or perhaps it has to do with composition of the tissue. As Keith was saying, differential cellular concentration perhaps. It could also have to do with how, so here the example of the dorsal interior cingulate cortex for how variable this region is across individuals.
Now I’m going to be presenting some new evidence of task-related behavioral change that we have just collected, we just finished collection a few weeks ago. So I will use this little example here, where you have two cues to choose from, called A and B, and you have no prediction over those two cubes, perform an action, observe an outcome, that triggers some sort of positive subjective feeling that can update your prediction over that QA.
These can be formalized using reinforcement learning models, which could define a prediction error. This is a quite famous signal in the learning field, which will be simply the difference between the outcome and your expectation over A, which will be scaled by a factor and used to update your prediction over that cue. Simply how you make a decision will be the probability of choosing A again in the future will be the difference between the expected value of the two cues.
And the reason that I am talking about those models is they fit behavior extremely well. So this is just an example where you can see in black the behavior of a real participant in choosing certain cubes, in blue the model. I don’t have to go too much in detail. I will also say these models have been gold standard in the field of learning since the early ‘90s, where it was discovered that the dopaminergic cell, particularly in the midbrain, multiple nuclei in the midbrain seem to be mimicking that reward prediction error at time of outcome.
So in humans how you would be translating this, you would just be checking the amplitude of the prediction error over time, combine it with the hemodynamic response function, and you would get a network. And that network would be showing the nucleus accumbens as the main hub. So the idea here in this study was to neuromodulate prediction error signaling by stimulating this deep brain region, which is the nucleus accumbens.
So the experimental design is quite similar to the spectroscopy study in the sense that it is a repeated design, people came four times, the first time for a UTE and a T1 weighted image that was used for acoustic stimulation, but also for neural navigation, and then they came three more times where they performed the study after receiving either a unilateral sonications of the nucleus accumbens, or the active control with a dorsal ACC. And they could also have sham.
And on each day, they received the same protocol, so this was a five hertz patterned TUS protocol, before they went into the scanner, and then we collected a series of FMRI and spectroscopy images. And those, for each block here they were performing a simple learning task where they had to just learn the associations, the contingencies between cue and reward probability.
So to study the influence of prediction error I used a general linear model which is simply using the prediction error at time T, and choice stickiness, so to what extent you’re sticking to a certain choice, to a study if you’re likely to switch or stay on the next trial.
And here are the behavioral results. So at sham what you can observe, which is very typical, is that if you have been positively surprised by an outcome you’re more likely to switch choice and less likely to be sticky to whatever you’ve been doing, so you’re becoming a bit more exploratory.
And this was exactly the same after the active control, the dorsal interior cingulate cortex was stimulated. However, this was much more enhanced after the nucleus accumbens had been sonicated. Which is quite aligned with the hypothesis that the nucleus accumbens is encoding this reward prediction error, but in this case what we’re seeing is a faciliatory effect. And they also become more exploratory.
What was interesting with these studies is that all of a sudden we could look at this effect over time, and what we can see is that the effect is much more pronounced, about 30 to 40 minutes after the intervention, and seemed to cool down after 50 minutes. And unfortunately, I don’t have the fMRI results yet because we’ve literally just finished collecting data. And this was not the case if I was looking at the dorsal interior cingulate cortex.
Just to say that even though a wide parameter space has been used across studies, more studies in humans are showing effects in the hour following the intervention, and even sometimes days. And here I’m also presenting some nonhuman primate studies, which really pave the way toward what we are doing now in humans.
So those studies have shown local changes in BOLD, in perfusion, a bit like what Keith was talking about, remote effects including functional connectivity changes, but also behavioral changes. And in humans some of those have clinical implications, for example the alteration of mood after sonication of the dorsal lateral prefrontal cortex.
Now, to achieve long-term improvement, potentially treating brain disorder, we have to refine the relationship between dose and response, and we need to understand how many times the intervention will have to be repeated, perhaps spaced out weekly or monthly. And there is very little knowledge of this yet. So much more research is needed in this area to clarify this.
And so far, there are just a couple of papers that are showing long lasting change over the course of months. This includes two very important papers in nonhuman primates where they were sonicating for even up to six months, but also showing that there was no damage after repeated exposure to sonication. So very important papers. And in humans there is very limited evidence of full sample studies, but there are indications of some behavioral change, which is very promising.
So what are the bioeffects of such repeated treatment? We know from small animal studies but also in vitro that several mechanisms are likely to be at play. So TUS might trigger long lasting synaptic modifications through LTP and LTD. TUS may influence astrocytes mediating the synthesis and release of brain derived neurotrophic factor.
And there is this beautiful review by Blackmore et al which is summarizing a lot of the biomechanisms at play for acute but also potential long-term effects. And also mentioning the potential influence on neural system cells, and also on the extracellular matrix, which could allow for a more permissive environment to support cell proliferation, neurogenesis, and overall neural plasticity.
So on this I would like to conclude and reiterate that there is more and more evidence that TUS can induce delayed effects that are happening in the hour following intervention, which is very exciting when we are thinking about clinical implications and mental health issues, including psychiatric disorders. Thank you very much for your attention.
GRACE HWANG: I want to thank both our presenters for excellent talks. And I want to kick off the discussion session. So we have several questions from the audience. I’m going to start with a quick clarification Elsa on your result with respect to task related behavioral changes. There’s a question from Maureen Goodman. He says here I see that the pulse repetition frequencies were 8 hertz, what was the pulse width and duty cycle of TFAS?
ELSA FOURAGNAN: It was 20 millisecond pulse duration for 200 milliseconds, so 10 percent duty cycle. The pressure in the brain was about 0.6 megapascal, and on average the in situ ISBPA was maybe 8 per centimeter square.
GRACE HWANG: Similarly, I would like to move to a clarification for Keith Murphy from Joann (name). Why is the duration of the stim different for the excitatory and inhibitory protocol?
KEITH MURPHY: I think I wrote a response. But we took an unbiased approach where we expanded duration and lowered duty cycle, or shortened duration and increased duty cycle. And that is just what we found, inhibited cell types. So that’s just kind of where we ended up with that.
GRACE HWANG: Thank you. Now if I go back to Elsa I have a question here from Holly Lisanby for Elsa. What is your understanding of why the same TUS protocol on the same individual has different impacts on GABA in posterior versus interior cingulate?
ELSA FOURAGNAN: This is a very good question, and one we have tried to answer for a while now. So from our data we did look at the idea of state dependence by looking at the contribution of the posterior and anterior cingulate cortex to their respective networks, and relate this to the changes in GABA. And although that was going in the right direction it wasn’t significant. It wouldn’t allow us to state that from those data it was pure state dependence, but there was some sort of a trend effect there.
It is also possible that there are differences in target engagement. So PCC is generally more deeper than the ACC by maybe a centimeter, which is quite important, which would mean the focal volume might be a bit bigger. And maybe there are questions of cell specificity within the regions, and this is all about what Keith is doing.
GRACE HWANG: Great. Let me go back to Keith. We have a question here from Ovid Yagnazanday(ph.). What change of behavior did you observe in the rotarod test?
KEITH MURPHY: Thanks for that question. We had looked at other hypothalamic subregions, and instead of walking sort of normally, which we observe with the DMH, we get animals to do an anxiety type of walk called stretch attend, and we wanted to know which of those sort of overrides the natural willingness to walk on a rotarod. And it turns out that just the DMH, the walking that we’re inducing is sort of unnatural, and it actually impairs their ability to walk when incentivized. That’s coming up in hopefully a pre-print soon. Hopefully you all can download it.
GRACE HWANG: There’s a related question for you, Keith. Fascinating work from Amram Shelacky(ph.) can you elaborate on the local stimulation resolution with TUS? Is it different with the miniaturized setup for free behavior? And second, are there methods for cell type specific targeting (inaudible) that you’re able to elaborate on.
KEITH MURPHY: There are certainly more focused beams that you can have on a freely behaving animal. And you’ll hear about some of those later in the conference. For our device we needed it to be well integrated to the fiber. And our field width was about two millimeters wide and five millimeters long. So it can be better. We’re hoping to redesign the probe to make it much smaller. And of course as you saw with the PET imaging, it’s not necessarily about the field size but sort of a titration of effect, and you’ll get always the strongest effect at the peak of the focus typically.
GRACE HWANG: This leads me to a different question, for Keith and Elsa. So the photometry coupled ultrasound that you developed Keith appears to be a very versatile way to monitor immediate ultrasound effects from many different brain regions. However, it appears that an invasive fiber may be required to obtain an immediate readout. Could you comment on a direct path for clinical translation to human studies?
KEITH MURPHY: I will say that Elsa’s study is really exciting for that same reason. We are actually trying to work together to show in photometry that there are cell type differences, and hopefully translate that to humans. The fiber is actually just a way to read out activity from the brain. But it’s not needed. So for the PET study we actually removed the fiber entirely, and just stimulated the brain. But you need another way to image the brain if you don’t have a fiber. And I’ll let Elsa comment on the best methods for that. For humans, Elsa, what are some of the approaches we’re taking to image noninvasively?
ELSA FOURAGNAN: Which cells would you want?
KEITH MURPHY: Whatever cells you want. Maybe start with GABA.
ELSA FOURAGNAN: The thing with GABA with spectroscopy is that you are just measuring the concentrations of the metabolite. It’s not necessarily related to cells, it could just be the receptors GABA A and GABA B that are across multiple neurons or other sites. So I’m not too sure that spectroscopy would be the answer.
But if you were looking at glial cells for example, because you are interested and many others are interested in astrocytes and so on. So PET would be the way to go, but it’s perhaps not the greatest for healthy participants and longitudinal studies. But diffusion weighted imaging could be the way to go.
GRACE HWANG: I have a question for you both. Could you please comment on what new technologies and hardware for generating focused ultrasound you would like to have? Forward looking, wishful thinking.
ELSA FOURAGNAN: Sorry, Grace, I was reading the questions on the Q&A.
GRACE HWANG: Could you please comment on what new technologies and hardware for generating focused ultrasound you would like to have? Sort of creeping towards what are the future gaps, what are your current gaps and what technologies would you like to have?
KEITH MURPHY: I guess I will start. I would love to have very fast raster scanning around the brain. It’s much easier to do in animal models, of course, and many labs are already doing this. But the ability to know where the focus is and to move it along oddly shaped structures, or to intervene in two circuits almost simultaneously. That’s on my wish list.
ELSA FOURAGNAN: On my wish list, since I am very interested in the prefrontal cortex, which is quite uniquely interesting in humans, I would like to have multi-element RA to be able to go beyond the issue of the sinuses that we have with simpler transducer.
GRACE HWANG: Great. I guess a straightforward question from Wes Lewis is do you fabricate your own transducers.
KEITH MURPHY: I do, yes. We have a machine shop that makes individual components, and then we use epoxy to stack everything together. I think even for some human studies early on people were sort of fabricating their own from piezoelectrics, but now I think there are many options for human studies.
ELSA FOURAGNAN: There are a few, most plug and play systems that are used for research, but a lot more that are currently being developed, and surely the field in the next five years is going to change quite a lot, because there will be a lot more technology out there.
GRACE HWANG: A few more questions, mainly for Elsa, from the audience. We still have four minutes on the clock. What is the data on the duration of effects at the single sonication stimulus in humans?
ELSA FOURAGNAN: So a single sonication, would that mean a single pulse? Because what I was presenting earlier was just the result of a single protocol of 80 second. A lot of the work had been done in nonhuman primates I feel that are showing as well that just 40 second of a train of pulses can have long lasting effect for short to medium lasting effects. So what I was presenting is what I would say is the results of a single exposure to two assays, no repetition there. The participants came three times but they were receiving sham or sonication in other parts of the brain.
GRACE HWANG: Thank you for that clarification. And then another clarifying question is when you use the word low damage in your presentation, is it in consideration of thermal damage, or other physiological response?
ELSA FOURAGNAN: When I was talking about the wet study, that was histological observations. So yes, it would be thermal damage and cavitation.
GRACE HWANG: So there is a question from Doug Minicky(ph.). And he starts with, perhaps this will be discussed in more detail on the biomechanism section, and this is referring to your talk again Elsa. I believe during the DACC local and remote changes part of the presentation. So the question is what are the hypothesized molecular substrates involved?
ELSA FOURAGNAN: What we are observing is perhaps the results of LTP-LDD, is that the question? So what would be explaining an effect that lasted to an hour, and the release of GABA. Again, we need to think about what spectroscopy measures. This decrease of GABA is surely due to the fact that there is less GABA in the extracellular space. This is what has been asked. It indicates that there is an excitatory response, and that overall there is less GABA in the extracellular space, so more in the vesicle within the neurons, in the cells.
GRACE HWANG: Thank you so much. One last question about any data about threshold effects on white matter versus gray matter. That just came in from Steven Rasmussen. I believe that was directed to Elsa. If Keith, you have thoughts on that feel free to chime in.
KEITH MURPHY: I have not personally explicitly looked at the area, but I think Sean Francois Brun(ph.) may be talking about differences. They use the Insightec to highly precisely focus on white matter and grey matter, and I believe they saw some differences. So I think it’s an exciting new area that I hope people look at, especially from a biophysical perspective, these different types of tissue morphology might lead to differences in ultrasound sensitivity.
ELSA FOURAGNAN: We have been generally looking at focusing on the cortex, but there is also the area of the radiation force and everything that is going around. So it would be very interesting to look at what is happening in the white matter for sure.
GRACE HWANG: Thank you so much for this action-packed discussion, Q&A. At this time I would like to turn the session over to Lizzy.
Session 2. Biophysical Considerations
ELIZABETH ANKUDOWICH: Thanks Grace. Thank you for that fantastic session. Welcome everyone to our next session today, focused on biophysical considerations. For those of you who do not already know me, my name is Lizzy Abkudowich, I am a Program Officer in the Division of Translational Research at NIMH, where I run the multimodal therapeutics program. I have the pleasure today of introducing our speakers.
Up first we have Dr. Kim Butts Pauly from Stanford University who will be speaking on biophysical effects. She will be followed by Dr. Jeff Aubry from Physics for Medicine Paris who will be speaking on biophysical safety, and who is also our moderator for this session. Welcome, Kim, and feel free to share your screen, and thank you so much for joining us.
KIM BUTTS PAULY: The goal of my presentation today is to take a step back and to explain ultrasound a little bit from the ground up, and explain, sort of unpack some of the concepts that were discussed in the first session, the first talks, and then give us some groundwork for moving forward with the rest of the sessions. So I’m going to talk about physical effects and biophysical effects. And my disclosures are shown here, a number of companies that I am working with.
So we are all very familiar with sound waves, pressure waves. And if you’re hearing my voice right now you’re hearing a pressure wave. It is simply that there are oscillations of high pressure and low pressure that will go through tissue for example. And we see that the particle motion right there is very small.
What do I mean by particle? It could be like a small amount of tissue, that is kind of how the ultrasound field describes a particle. And what we see there is the energy is passed from one particle to another, and the wave will past much more quickly and longer extent than the particles themselves.
So like I said we’re all familiar with that. You take your tuning fork, and you can ping your tuning fork, and then what we will see is these pressure waves will oscillate, they’ll go into the external auditory canal, impinge on the ear drum, go through those three little bones and then be conducted over to the cochlea.
Now, what do I mean by frequency? Well if you’re standing at the ear drum and you have a stopwatch and you’re counting how many high-pressure points per second, then you’re counting the frequency. We humans are most sensitive to frequencies between 32 hertz and 16 kilohertz. So those are audible frequencies for air conducted sound. So we define ultrasound as frequencies that are above that for air conducted sound, as in those frequencies above 16 kilohertz if it’s air conducted we don’t hear those. We’ll get back a little bit later about audibility of ultrasound.
So what we’re going to do is we’re going to change our transducer from that tuning fork to something like a piezoelectric crystal, and it’s going to create those pressure waves that are going to go across the skull and then create pressure waves in our tissue.
So here is a simulation of a pressure wave. This is using a curved transducer. And usually we’ll be talking about curved transducers. So we’re going to focus the ultrasound. It’s called focused ultrasound. And what we see are those pressure waves going through the tissue.
So there are oscillations of high pressure and low pressure. And if we zoom in on the center part right here and we look at one particle, we see that it has some forward and backward motion, it gets compressed and stretched, stretched and compressed over and over again. And with each cycle of the ultrasound. And each cycle lasts on the order of microseconds.
So what we see there is that stretch and compress, it undergoes what is called a normal strain. That’s in the direction of the ultrasound beam, it gets stretched and compressed. Strain is a deformation, it’s a unitless quantity, it’s given in a length ratio, often given in parts per million. So there’s normal strain. And then there’s a shear strain, as you look at those particles on the side of the focus, you can see it’s kind of like a rubber sheeting of the tissue. And so that is a shear strain.
So if we want to ask the question well what are we talking about in terms of numbers, the normal strain first is just a very gentle oscillation or deformation of that tissue. We did a simulation where we were looking at a transducer that is going to give us a nice focus, so it’s a curved transducer there. We were looking at a peak pressure of one megapascal. It’s a nice, round number that we’re looking at. You can see in red it goes up to one megapascal, a focus around 30.
What we see for the deformations then or the strain is about 800 parts per million. What does that mean? Well, if you think in terms of the millimeters, the axons are lasting over a millimeter, something the size of a cortex is a millimeter, we’re talking about a deformation of about 0.8 microns. If you’re thinking about maybe the size of the cell body, of 10 microns, we’re talking about eight nanometers. So it’s a very small oscillation that we’re talking about.
And then there is the shear strain. So again, it is sort of a rubber sheeting of the tissue itself. And then the types of numbers that we are looking at are about a quarter of what we were seeing for those normal strains. So these were some simulations that we were looking at at a half megahertz.
So particle motion strain is one source of strain. There is also another source of strain, something else that’s going on, and it’s the acoustic radiation force strain. So that was mentioned before. And I’m going to start by giving you an illustration to describe what I mean here.
And in my illustration, it’s the child on the swing because it’s something we’re very familiar with. The parent is going to push the child on the swing, put energy into the system. And in the absence of any losses what we see is the child is going to go forward and backward and they’re going to oscillate around the zero position straight down into the earth. However, if we do have losses, friction and wind resistance, then what we see is the child is going to go forward higher, and then when they come backward, because there are losses over time.
And what we can see now is there is a net displacement forward. In many ways this is analogous to what we see with the acoustic radiation force displacement. So with ultrasound there is going to be some loss of energy as the pulse is moving through the tissue, it’s absorbed as heat. And then there is going to be a net displacement forward, and then that is going to give us a strain much like the particle motion gave us strain.
Our simulations here give us an idea of where we see those types of strains. We don’t see the strain so much at the focus of the acoustic radiation force, we see it in front and behind, and we see these shear strains on either side. In these simulations they were a little bit lower than the particle motion strain.
But one of the points that I really wanted to get across to everybody today is that with ultrasound there are a number of things that are going on simultaneously, and we can never eliminate all of them, we can never really eliminate any of them at any one time. Sometimes we can play up parameters, so that we can play up one of these effects versus another. And so for example you might change frequency, that might be able to change the acoustic radiation force strain versus the particle motion strain.
But let’s step back for a moment. I know what you’re thinking. You’re thinking at this point, Kim, this is really straining my brain to talk about all these strains. But there are really good reasons to talk about this. And one of the reasons is this work by Scott Hansen that really resonates with me. It’s because his work has been looking at the fact that mechanical force acts on lipid membranes.
So what’s going on here is first of all, to give you a backdrop here, is that lipid membranes are made of different types of lipids. And so there are lipids that are more ordered, and lipids that are less ordered, or disordered. Sometimes he says to me think about it like lard floating around in olive oil for example. In the order domain there are molecules that reside in the order domains.
And then some kind of forces, so we’ll have a mechanical force, a stretch or a shear, then those molecules will come out, because the ordered lipids will become a little bit more liquidy, and such that those molecules can come out, they’ll translocate, there’s this whole cascade of events that occurs, leading to the activation of stretch sensitive ion channels.
He did his work looking at TREK-1, this is a potassium channel. We know that other channels are also mechanical sensitive, the TRP channels for example also have this signaling cascade. So super fascinating to think about how it’s really the lipid membrane that senses mechanical force. But all of these effects that we would talk about in terms of strain are really relevant here.
The other thing as I mention is that there are a number of things that are typically going on with ultrasound at the same time. And pressure and temperature, I’ll talk a little bit more about temperature in a moment, these will also have an effect on the ordered domains.
So it's a little bit hard to sort of separate out, when we apply ultrasound, to say all of the effects are necessarily from this physical effect versus that one, except as I mentioned we can sort of play with our parameters a little bit, so if we can get down to a temperature that is very low then we might be able to say it’s more likely to be due to the mechanical effects, and et cetera.
But again, what I wanted to do today is setup the scene here for the different types of physical effects that are going on. So I’ve talked a bit about pressure. I’ve talked about strain with its two different types of strain. And let me now tell you a little bit about temperature.
Before I talk to temperature I have to step back for a moment and define intensity. So you’ve heard intensity and pressure a little bit, and so let me define that for you. If you have an ultrasound pulse, and it’s oscillating at the frequency, the fundamental frequency, maybe that’s 500 kilohertz. You have an envelope, you can look at the envelope and say there is a peak positive pressure or peak compression pressure, there is a peak refraction or peak negative pressure.
Now we’re going to define the intensity. You simply take that pressure, and then you square it and divide by a constant which is specific to the tissue. So it’s called the acoustic impedance, it’s a tissue relevant parameter, and it is defined as the density times the speed of sound. So then you can get the intensity. And it might look something like this.
Well, given that we have a pressure, we can calculate our intensity, some other parameters we might be interested in maybe is what is the temporal peak of the intensity, what is the pulse average of the intensity, or maybe we’re interested in the temporal average, which takes into account the ultrasound pulse itself and then the dead time afterwards.
Let’s put some numbers to this. If for example you have a peak pressure of one megapascal, we might then be talking about a temporal peak in brain tissue about 60 watts per centimeter squared, and a pulse average about 30 watts per centimeter squared.
Some assumptions that went into that, this just gives you some ballpark numbers. Unfortunately, it is a little hard when you are listening to talks to kind of convert in your mind between pressure and intensity. And so that’s why I wanted to give you these numbers to kind of give you just an idea of what we’re talking about here.
So as I mentioned you’re going to hear talks from people talking about pressure, and some talking about intensity. And why do we talk about both? Why do we not just settle on one? And it is because different ultrasound effects will rely or depend on one versus the other.
So we talked about particle displacement, which is dependent on the pressure itself, whereas temperature and radiation force are dependent on the absorbed power, which is related then to the intensity, and that’s the pressure squared. Unfortunately, we’re just going to end up talking back and forth between those two parameters, but at least now I’ve defined it so we know what to say going forward.
Now, temperature. We do know that ultrasound can induce temperature. We know this from the essential tremor treatments, we use a large area transducer, we focus to the VIM nucleus, we can get an MR temperature map, and then from there we can look at that peak temperature there. We see that while the ultrasound is on the temperature rises exponentially, and when you turn it off it decreases, and it decreases exponentially over many seconds, and we can treat those patients.
But what I wanted to talk about here is this plot right here and what governs this. So it’s governed by an equation shown here, which is a little complicated, and I’ve made the text purposefully really small so that you don’t read it. I don’t want you to look at that text.
What I want to tell you, and what I want you to take away, is that there is a heat change term here, so that’s what’s going on with that curve up there. It’s related to the heat deposition term. So this is how much power you’re putting into the tissue, the power then is the intensity, and then there is the absorption coefficient.
Then there are two heat dissipation terms. The heat conduction term, so that’s simply if you have a spatial gradient in the temperature, the heat is going to go from high temperature to low temperature. And then lastly a perfusion term. So when the ultrasound is on, over short times, it’s then typically going to be governed by this heat deposition term. When you turn it off it’s these two terms over here that govern it. And then you can find the temperature rise.
So now the question is well when we think about what we’re doing with transcranial ultrasound, how is that now related to maybe sort of the power levels that we saw with the central tremor that cause these temperature rises that you see in that plot. And just sort of ballpark numbers, what we see with the central tremor is that we’re giving typically something like 100 watts per centimeter squared for about 10 seconds. So we’re depositing about 3000 joules.
And here with transcranial ultrasound it’s much lower. Maybe we go up to one megapascal, which I said was a pulse average of 30 watts per centimeter squared. Maybe it’s about 100 milliseconds for that pulse. That’s about three joules. If we wait some number of seconds, then we’re sort of starting over with that temperature rise, because that decrease is going to happen over a couple of seconds. So much smaller, by three orders of magnitude typically.
That’s where we are with temperature. Let me now introduce the idea of cavitation. Cavitation unfortunately means many things. The one that is most relevant to us here, well there are really two that are relevant. First of all, it relies on the fact that there are gasses that are dissolved in our tissues. We know we have nitrogen, we have oxygen, carbon dioxide, et cetera. So we have gasses dissolved through our tissue.
And when we apply ultrasound, during that low pressure part, where there is low pressure, then what we do is we can pull those gasses out of the tissue and they form a bubble. Then the ultrasound is going to go through that high pressure phase, the bubble is going to compress a little bit, and then it is going to go now back through its negative pressure, low pressure phase, and the bubble is going to expand.
But it’s not only just going to expand, but it’s going to grow. It’s going to grow because the concentration of the gas in the tissue is going to be higher than in the bubble. And so as that bubble expands it’s going to have low concentration of the gas, the gas is going to diffuse into the bubble.
A couple of things that help that bubble growth. One is that as the bubble expands the surface area gets really big. So there’s large surface area for that diffusion. And then the contraction part, there’s very small surface area. So we’re really going to be preferentially having those bubbles expand.
And then lastly, this is a movie given to me by Larry Crum that shows the bubble expansion, and that it spends more time on expansion than it does on contraction. So that contributes to what is called nonlinearity. So all those together mean that we could have the possibility for bubble growth. We’re not necessarily going to see it, and that is going to depend on our parameters.
And so all of these parameters, all of these things, these effects are going on simultaneously. They could all play a role in the stimulation that we see. It’s a little hard as I said to sort them out. And what we’re going to see in session three is a larger discussion about the biomechanisms.
All of these could be taken to an extreme, and if taken to an extreme then they can have an effect on safety, and be less safe. So what we’ll see in the next lecture is how do we think about safety, what do we know about safety, where is the boundary, and et cetera. And that is going to be a large discussion point with the parameters that we’ve seen.
We’ll talk about this more as well, what we’re seeing is that these pressures that we’re seeing that are under the FDA regulatory limits right now, that we’re very gently having mechanical force on the tissue and creating these little strains and affecting mechanical sensitive ion channels. And again, you’ll hear more about this.
So, where are we? At this point, the promise of transcranial ultrasound is that we can use external transducers, we can focus deep in the brain with small focal spots, and gently perturb these mechanical sensitive ion channels. Along the way there are going to be some speed bumps. And one of the speed bumps right off the bat is the skull.
So I’d like to spend a bit of time talking about the skull, and kind of what are those physical attributes of the skull that make it difficult for us. One of them is that there is variability in the skull, there is attenuation, and there is speed of sound. So let me explain these in a little bit more detail.
First of all, let’s look at the variability. Here is a slide with a number of skulls. These are subjects that were treated for essential tremor, so a heavy CT, and what you see is there is a huge heterogeneity between and within individuals. So you can see that you might have very thin skulls, very thick skulls, thick skulls that are thick cortical bone and thick trabecular bone. Typically the skull has three layers, an outer layer of cortical bone, an inner layer of cortical bone, and in between there’s trabecular bone.
And the skull is highly attenuating, which means that as we apply ultrasound there is going to be a loss in the pressure amplitude. There is going to be a loss of the intensity amplitude. What we see is that there is a big loss in the cortical bone, but it’s even bigger in trabecular bone.
So if we look at these and we say which is the most attenuating to ultrasound, there’s a tendency to look at this skull over here because it’s so thick and it’s so white, you key in on that. But the truth is it’s the second one that is the most attenuating, because it has so much trabecular bone right there.
Okay, variability is a problem, and attenuation is a problem. Let’s continue with a story, because there is a lot going on here. Speed of sound is different between water or watery type materials and bone. So we can see that here, if you can see that high pressure point, it goes through the bone more quickly than it does through the water in this animation. If we look at the numbers we see that for water, aqueous soft tissues, it’s around 1500 meters per second, and in bone it’s about twice as much as that. Why is the speed higher? Well mainly it has to do with the compressibility.
So something like air for example is very compressible. So you can press on it and the molecules are going to go very far before they impart that energy to the next set of molecules. Water is in between. Bone doesn’t go very far before it sends the energy to the next set of molecules. So compressibility is the overwhelming factor here. Low compressibility in bone means very high speed of sound.
So, high speed of sound in bone as well as high density. So this number that I talked to you about before, the acoustic impedance, that’s the density times the speed of sound. And in bone what we see is both those numbers are very high, giving us a high acoustic impedance.
With air, both those numbers are very low, giving us a very low acoustic impedance. Okay, what does that mean for us? Let’s say we have a situation where we have our transducer, we have our coupling material, and then we have a little bit of an air bubble before it hits the scalp. What we’re going to see now is there is going to be complete reflection at that interface. The reflection is given by this equation here.
And when there’s big differences in the acoustic impedance, such as between air and soft tissue, big differences there mean big reflections. And so we’re going to see a big reflection from any kind of air. This is why we couple the ultrasound to the skin with a coupling material, we’re very careful about that, and this is why we worry about air bubbles in the hair.
Well, what if we did a good job, we’ve got our coupling material, and now what we have and we’re thinking about is scalp, and we’ve got our skull, and then brain. Again, what we are going to see is big differences in the acoustic impedance, we’re going to have a big reflection. We’re going to have a reflection at that first interface, a reflection at that second interface, and it's possible these other interfaces in there as well, but definitely those two interfaces we’re going to see reflections.
So, how much at each interface? Well, you could put in the numbers if you want to. You’re going to get something like 50 percent of the intensity is transmitted at each interface. And again, this is very on proximate. But that means now two interfaces, it’s going to be 0.5 times 0.5. This is pretty significant.
So now let’s consider three cases of constant thickness. Let’s say we have thin cortical bone, thick cortical bone, and thick trabecular bone. First of all, we’re going to do a good job of our coupling materials, we’ve already covered that importance.
If all we have is attenuation, maybe we put in the attenuation numbers for soft tissue, then we get something like 60 percent of the intensity at the focus. But of course that’s not going to be true, because then we have to take into account the loss at the interfaces of the bone. Okay, now we see that we’re getting down in how much intensity we’re going to get at our focus.
But then if we put in the attenuation as well it gets even smaller. So that thin cortical bone was beautiful, we get 50 percent, and it could be quite a bit smaller with a thick trabecular bone. And that’s something we know from the central tremor pre-treatment, that some people are hard to treat if they have too much thick trabecular bone.
Well, this was sort of a nice ideal situation where you have constant bone thickness. And so now the question is what if your bone isn’t a constant thickness, now what is going on. So the idea here, you’ll see this in this animation, is if you have your transducer over here, and it’s sending the pressure wave into your scalp, which I called water, and now it’s going into the bone of different thicknesses, because it speeds up in the bone it’s going to come out at the other side.
And in any one position now you can see that those different parts of the ultrasound beam are no longer in the same phase. So where there’s high pressure up here, there’s low pressure there. So there’s going to be destructive interference. Those different parts of the ultrasound beam are going to be incoherent, and you’re going to have a loss of signal. This is what’s happening in this set of simulations, where we have a nice focus, we have more aberrations, and we get to something like that.
What does it look like on these skulls? Well these over here you might say we’re going to have sort of a medium level of aberration. This one over here, you put it at this spot, it’s going to be a lot of aberration, you’re really not going to get much at all at your focus.
And what’s really interesting also is if you think about what is happening over here, you might get, if you put your transducer here, a really nice focal spot. But it may not be exactly where you want it to be. And this is because there is that variable thickness of that skull, it’s not giving you a complete aberration, but it’s basically bending the wavefront because of that high speed of sound through the skull there. And this is called refraction. So where you want there to be a nice focus isn’t there but it’s somewhere else.
Well we will have reflection and attenuation. We might have aberrations. And we first of all absolutely need to assess this on a subject specific basis. So this is going to be why we’re going to talk in a whole session about modeling and being very rigorous about our understanding of the focus.
Another question you might ask is can we get a measure of the focal pressure. Well, I’ve already mentioned the acoustic radiation force. This is one possible way that we can get a picture or measure of the focus. The other one was of course temperature, and I’ve shown you a picture of a temperature map as well. With temperature imaging we could probably get temperature imaging that’s reasonably good at about a degree C.
We would struggle to get a signal to noise of less than a quarter of a degree C. but what about acoustic radiation force? So we already talked about that, where what is basically happening is if you’re looking over timeframes of milliseconds then all the oscillations of the fundamental frequency are averaged out, and you’re just looking at that big displacement from the radiation force.
With MRI we could set up a pulse sequence that’s sensitive to that. The pulse sequence looks like this. It’s a spin echo, so we measure a signal over here, we’re going to motion encoding gradients, so this looks a lot like a diffusion weighted sequence. We’re going to turn our ultrasound on for one of those motion encoding gradients.
And then what we’ll see is the spins will move, now we’re talking about water spins, they’re going to move to a higher magnetic field, and they’re going to spin up-phase because they are in a higher magnetic field. So basically what we have done is we have encoded into the phase of the image what that displacement was.
And here what we see now is the focal spot, this is an ex vivo porcine brain, we looked at the focal spot. MR-ARFI then can tell us the ultrasound is on, it’s working, we get a nice focal spot, it’s in the position that we want it to. And now the question is can we back out of that picture something more quantitative about our ultrasound beam.
We looked at that in sheep. So here is an MR-ARFI picture, in vivo, a live sheep, and we’re able to say again we get a nice focal spot exactly where we want it to be. In this experiment we were sonicating the LGN and looking at the visually evoked potential, and we see a reduction in visually evoked potential, as you can see on this scale. What we saw is that on the horizontal axis we looked at the displacement from the MR-ARFI.
And over here there is a lot of displacement that was a situation where there is low skull attenuation, so a lot of intensity got into the focus, and that’s where we had a bigger neuromodulatory effect. Over here there was higher skull attenuation and a lower neuromodulatory effect. So not only can we see that in fact the focal spot looks good, it’s where we want it to be, but maybe we can say something prospectively about what kind of an effect we’re going to have in terms of the neuromodulation.
So, can we correct for those phase aberrations? If you have a multi-element array you could. So the idea here is you could do something on those elements. You could have a time delay that’s a little different. It’s easier just to put a phase term on those, but the effect is the same. Now what the simulation shows is that when they come out after the bone right there, now each of those little wavefronts is now in-phase, so we’re going to have a better focus.
So this is what the essential tremor treatments do, they have phased arrays, and they can correct for phase aberrations. And it does require phased array. Some of the systems that people are using right now are single elements that don’t let you do that, or there may be angular arrays. An angular array will let you have some depth focusing, but it doesn’t give you the ability to correct for these phase aberrations.
So where are we? We will have reflections and attenuations; we will have phase aberrations. And we could correct for phase aberrations, if we can measure and model them and have phased arrays. So that is definitely where the field is going. You’re going to hear more about this in the modeling as well.
I have just a few minutes left. And one of the things I wanted to queue up as well is not so much what’s happening with the tissue from the ultrasound in terms of the neuromodulatory effect, but I wanted to tell you a little bit about sort of the physics of where these auditory effects come from that you’re going to be hearing about. So I’m just going to tell you a little bit on the basic concepts, and then later in the workshop you’re going to hear a little bit more from people about efforts to mitigate the auditory confound.
So where does it come from? Because what I already said is if you apply ultrasound, it’s done at frequencies that are defined as ultrasound, meaning they’re higher than the hearing range for humans for air conducted sound. However, that can still be heard by humans. And it’s simply, there’s multiple pathways by which the pressure wave can then come in and be heard.
So the sound pathways are shown here. First of all, the sound can come into the external auditory canal, and then pass through the tympanic membrane, the normal pathway. It can come through the temporal bone and into the cochlear directly. Or it can come via the soft tissues and through the fluids in the cochlear aqueduct. And so that’s called the third window.
So there are multiple pathways here. And ultrasound can be heard. And the reason it is heard is not because of that fundamental frequency, but it’s because of the envelope. So that’s what is shown here, is the envelope. So again what we have is a pulse that is oscillating at very high frequency, say 500 kilohertz, but we’re going to turn it on and turn it off for some time.
And in this example, it’s 80 milliseconds long. Maybe you’ll hear about pulses that are one millisecond long, or different sets of durations. Inevitably there’s an envelope here, and there’s an oscillation of the frequency. It’s this envelope which is most audible to people. If you look at the power spectrum, so this is the frequencies here, this is the human hearing range, from 32 hertz to 16 kilohertz, then this waveform has large frequency components across the human hearing range.
This is the mouse hearing range here. So they have higher frequencies that they hear. We did do a study where we looked at this waveform, we looked at an ABR, so this is putting electrodes at the back of the head and measuring with the EEG what are the signals coming from the cochlear up through the brainstem and into the brain.
And you see these squiggles here, meaning this was audible to the mice. It’s really these sharp edges that give you these high frequencies that are really very audible. Because then what we did was we smoothed those edges, and those high frequencies were decreased by 50 db. Now, remember every 10 DB is a factor of 10, so this is down by 100,000 in terms of the intensity. And now what we see is that the mice didn’t have that ABR response.
Now, that’s great, but unfortunately one of the difficulties of doing mouse research is that you can’t ask the mouse what they perceive. And so the mouse isn’t going to tell you well there was some residual sound that I heard even though I didn’t have an ABR. So we did do a human pilot, some work I did in conjunction with Len Verhagen and Ben Kop at the Donders Institute.
And here we did something similar to what we did with the mouse. We have a rectangular envelope for 20 milliseconds. We did three in a row, and we compared it to a smoothed waveform. Very similar to what you just saw. The frequencies that we’re most sensitive to, here now looking at two kilohertz or down by 50 db.
And then what we did was we asked the subjects as we played these two waveforms back and forth to tell us when did they perceive it, you change the amplitude back and forth until you find the threshold, plot the threshold here, and for all subjects there was an increase in the threshold, meaning this one was less audible.
That’s great, it was significantly less audible, factor of four in pressure. What’s really interesting as well is if you ask these subjects what did they perceive, this one, there’s three pulses in a row, it sounded like a tch-tch-tch, very complex sounds, exactly what we would expect from the frequency spectrum.
This one down here, we thought all of that was going to go away. And in fact, all of those complex sounds went away, which really corroborated our theory here about the envelope. But it did leave a high pitch that was not described here by the envelope. We need to do a little bit further work about this, there is more that’s known from the bone conduction literature. There’s multiple pathways that I mentioned. And so it’s possible that we’ve affected one of those pathways and not another one. And so what we do need to understand as well with humans, so what is that residual pitch, what is the frequency, and then how do we mask that as well.
I want to conclude at this point. What I’ve shown is that at these relatively low pressures and durations, and what we’ll talk about is that most of the work is under the FDA regulatory limits for diagnostic ultrasound, that these relatively low pressures and durations ultrasound can have physical effect on tissue, including strain from particle motion and acoustic radiation force. There’s also temperature and cavitation effects, and you’re going to hear more about these in a number of these different talks.
Any of these physical effects could be taken to an extreme and be damaging. And so Jean-Francois will talk in the next talk about discussing the safety aspects of ultrasound. A key challenge for us going forward, and something to think about as one of our key challenges for the session today is to discuss finding that therapeutic window where we can sort of maximize our efficacy while minimizing the safety aspects.
And then ongoing challenges include the skull, auditory confounds. Other things I wasn’t able to talk about was somatosensory confounds, choosing parameters, defining dose and dose rate. And then Tulik Nandi will have some discussion about parameters and dose and dose rate in her presentation a little bit later as well. So I would just like to acknowledge my group, my funding, my collaborators, and at this point turn it over to Jean-Francois for the next talk.
JEAN-FRANCOIS AUBRY: So I’m going to discuss the biophysical safety of transcranial ultrasonic stimulation. And we start with disclosures. I received grant from FUS Mobile company and also from Insighttex, ultrasound brain therapy. I hold stock options of FUS Mobile. And I am co-inventor of five patents in the field of biomedical ultrasound, for which I received royalties in the past, and I might receive some also.
Let’s jump into what we are going to discuss here, is what are the risks that are potentially associated with the use of ultrasound in the brain. The first risk is mechanical damage. And as Kim mentioned, this might happen when a bubble is present in the medium or it’s triggered by ultrasound, and it will oscillate, it will grow due to rectified diffusion as Kim explained, and finally it might explode, and it might induce tissue damage.
It is currently used actually for therapy, and this is how actually kidney stones are destroyed. In order to get rid of kidney stones they are crushed into small pieces thanks to cavitation. So of course this is something we would like to avoid in the brain definitely. So this is the first risk, mechanical damage.
It is also possible to induce thermal damage. And again, this is also used in the clinics to treat patients. Kim mentioned the use of the Insightec neurosystem in order to ablate tissues in the brain and treat essential tremor with this approach. And in that case, for thermal damage what is important is not only the temperature itself, you can see here it’s a question of temperature but also exposure time. So for a given temperature it might take longer or shorter to induce a lesion, and I will explain this in more detail later on.
But let’s start with mechanical safety limit, and what are the safety limits that we discuss. We can go to extreme levels in order to trigger this. So what are the limits to induce cavitation in tissues? I’ve looked back in the past on published data, and very few experimental studies have been conducted in vivo, trying to determine the threshold, to see actually cavitation, and there’s actually fewer studies in the brain, but there are some.
The first one is actually from a group in Paris. With Gateau we explore the probability of cavitation in sheep brain tissues. We used very high framerate ultrasound imaging in order to be able to detect bubbles, the tiniest bubble we could detect in the brain.
And we ramped up the ultrasound until we could see some cavitation activity. And the threshold that we measures was actually almost minus 13 megapascal, and we were using 0.66-megahertz ultrasound with very short ultrasound impulses. So very similar to what is used for diagnostic ultrasound, but of course with a much higher amplitude.
Other groups also were trying to take advantage of this therapeutic effect, not for kidney stones, but in tissues. In this case it is calls histotripsy, to actually destroy tissues by liquifying the tissues with ultrasound. This was a Michigan group.
And they looked also for thresholds to induce cavitation because they wanted to use cavitation to treat. And they also came up with quite high thresholds in that case, again with very short ultrasound impulses, minus 13 megapascal at one megahertz, and minus 21 megapascal at 0.75 megahertz. And those are either in (inaudible) or in tissues.
Finally, K Hynynen also a long time ago measured actually the threshold, this time for long sonications, and in that case for one second sonications he found a linear relationship actually in terms of threshold. The threshold was 0.6 megapascal plus five times the frequency being expressed in megahertz. So based on those data, two comments here. First, cavitation, the occurrence of cavitation depends on the frequency that we’re using. It is much easier to cavitate at low frequency than high frequency.
And also it depends on the duration of the sonication. For very short sonications, one microsecond, it is very difficult to induce a lesion and to see cavitation occurring in tissues, whereas the threshold is much lower when using long sonications. We will discuss this more in the second part of my talk.
So, one approach, if we would like to have kind of a table to know the safety limits, it means that those investigations would have to explore a large span of frequency and a large set of ultrasound duration parameters.
Now, switching to thermal safety limits, what has been introduced is the concept of thermal dose. As I mentioned earlier, in order to be able to induce thermal necrosis in the tissue, we need either to increase the temperature a lot for a short period of time, or with lower temperature we need to wait longer until we’re able to see the damage. And of course if we go above 100 degrees we will induce tissue boiling. But here what is interesting is to look at the threshold, to induce a lesion.
And the thermal dose is a very interesting concept. In this plot here, Sapareto and Dewey introduced the thermal dose, actually compare the set of temperature profiles, so temperature as a function of time. And actually what we see here is that by using any of those very different profiles the effect on tissues would be exactly the same. And they came up with what is called the thermal dose, which is the cumulative equivalent minutes.
So whatever is the time course of the temperature, the temperature as a function of time, if you calculate this, which is the integral over time, not only the integral of these curves, not the area under those curves, it’s a little more complicated than that, it includes here a constant coefficient that varies with temperature.
But they came up with this formula. And whatever the course of the temperature is over time, if you end up with the same number of cumulative equivalent minutes, which is the number of minutes equivalent to a temperature of 43 degrees, you will have the exact same effect on tissues.
How well does it work? Dewey and Arrhenius in 2009 in a clonogenic assay took CHO cells, and they exposed the cells to various temperatures, and for various times of immersions at those temperature. And you see the survival curves.
So the survival curves differ quite a lot depending on the temperature of the cells and the time during which temperature was applied to the sets. And you see here that the survival is decreasing of course for all of them, decreasing much more rapidly for high temperature, and much slower for low temperature. Now what they did is they calculated the thermal dose for each of these points.
And when you do so actually all those points align almost perfectly on the line here. And this is where it proves actually that this theory of thermal dose is actually very relevant to assess the damage to tissues and the number of surviving cells exposed to a given temperature. So again, this is a combination of temperature and time. So this is the concept of thermal dose.
Now, what is the thermal safety limit in the brain? And what is the minimum thermal dose required to see damage in the brain? And here in this paper from van Rhoon and colleagues, actually looked at several thermal dose thresholds as potential guide for MR exposure levels. And they looked at the literature, and what they found, and this is also the only paper I found with such a low thermal dose threshold, is that it is possible to induce tissue damage in the brain by applying temperature of 43 degrees for 7.5 minutes.
So 7.5 would be the threshold for inducing a lesion in the brain by increasing the temperature. So whatever, again, temperature curve you’re using, if you use the thermal dose calculation, if it’s below 7.5 degrees, you’re very unlikely to induce thermal damage in the brain, but above it might happen. One has to say here that this is the lowest value I’ve ever seen published, and apparently dog tissues are more sensitive than other types of brain tissues.
But it gives the limits in terms of temperature condition that can be applied to the brain. We can discuss during the discussion if anyone knows about anyone who would produce this work on measurements. It can be also questionable if this is the lowest value for thermal tissue damage, or somehow if we have more sensitive ways to assess the tissue damage, there might already be an effect for values that are lower than that, and that by the way, how can we reassess tissue damage. So in that case (inaudible) but I believe we will start to discuss potential damage in the brain.
So what I explained before for both thermal safety and mechanical safety is a way to assess it is just to look at thresholds at which above the threshold we can see lesions, and also ways to use transcranial ultrasound stimulation in a safe way, is actually to try to define safe parameters based on guidelines, on existing guidelines that have been actually defined for other medical devices.
So what are the closest guidelines to a transcranial ultrasound stimulation? Because there are no guidelines currently for transcranial ultrasound stimulation. Of course, imaging diagnostic ultrasound is the closest guidelines that we have. And also, the guidelines for magnetic resonance imaging.
So for assessing the mechanical risks, actually as I discussed before mechanical risks are actually due to the high pressure that we apply to tissues. And what is important is not the peak positive pressure, but the peak negative pressure. And the mechanical index has been introduced for diagnostic ultrasound systems, in order to assess the mechanical risk.
Here is the definition of the mechanical index for diagnostic ultrasound. See that it is the peak negative pressure here divided by the square root of the frequency that is used. So I already discussed before that the risk for cavitation is higher for lower frequencies.
So you see here when using a low frequency this increases the mechanical index. Again, this makes sense, of course it does. This is a way to assess the risk of cavitation for short impulses, because this is for diagnostic ultrasound. And currently recommendations are to limit the mechanical index in the brain, again for imaging to values that are lower than 1.9.
So, how does it compare to what I explained before, and what has been measured before? So in the experiment by Gateau et al, I showed you the peak negative values before. So the equivalent MI here would be 15. So way above the 1.9. And here the equivalent MI would be 13 and 24. This is almost the real MI because it’s very short sonications.
And with the experiments from K Hynynen, I should not be allowed to talk about MI because it’s long sonications, but still we’ll say the equivalent MI, just calculating the peak negative pressure divided by the square root of the frequency, the lowest MI would be 3.7 at 0.25 megahertz.
So I do believe that for mechanical index values lower than 1.9, even when using longer sonications, which was not what was intended when the mechanical index was first introduced, there seem to be no evidence of cavitation or damage due to peak negative pressure if the MI is less than 1.9.
You can see here that in this set of experiments we’re getting close to this value. And in that case cavitation has been seen. So we have to be careful again. If one would like to go higher than 1.9 value for mechanical index, it should be supported by experiments using the same frequency and the same duration and look at damages. It’s very difficult for results given at one given frequency with short impulses to compare with what would be obtained with longer things. It should be safe if the MI is less than 1.9, but for transcranial ultrasound stimulation, we are in a situation that is slightly different from the true definition of the mechanical index. To go back to the definition, so here the mechanical index is the peak negative pressure divided by the square root of frequency. But the peak negative pressure at the location where the pulse intensity is at maximum. And this value is derated by the attenuation of tissues.
In all cases for transcranial ultrasound stimulation, one has to carefully look at the location where the peak negative pressure is at maximum, because depending on the location of this curve, this point might be inside the brain, or outside the brain. And we don’t want to damage the brain for sure, but we don’t want to damage the skin either. So this depends on the setup, and it remains to be determined.
But also, especially in the case where the peak negative pressure is maximum here inside the skull, in the definition of the MI it is derated by soft tissues, but that is not the case, we know that the impact of the skull is much higher, so that actually we should be able to increase the power because the transmission through the skull is higher.
So in order to develop a conservative estimation of the transmitted power through the skull, we are as a group in the Attali consortium proposed a three-layer model, a simple model to assess actually the maximum transmission through the skull.
And basically, what we did is that we took a set of human skulls and we registered the skulls, we extracted the skull thickness from median points, and we data mine the thickness of the skull everywhere around the skull where ultrasound could actually got through for TUS. And from those data, we estimated the maximum transmission through the skull for different frequencies. And we know also we see here that it is very heterogeneous.
So the average transmission that can be achieved by the transducer is actually averaged over the surface area that is actually going through the skull. And so you see that the maximum transmission through the skull depends both on the frequency that is used and the surface area of the transducer, more precisely the surface area of the beam intersecting the skull.
And we provided here conservative values of transmissions that are actually the transmission is lower than what you would expect by just considering soft tissues. But of course, this is highly conservative, and we took here the maximum transmission that we found on millions of points on a set of 15 I think human skulls. So this is just to help assessing the mechanical index in the brain in I will say the most realistic yet conservative case, in which we take into account the effect of the skull.
In terms of thermal safety, there exists also some standards that we could use, because thermal safety effects are not limited to the effects of ultrasound. And for example MR devices are allowed to operate at a maximum temperature of tissues that are below 39 degrees, those are the parameters.
And for implantable devices it is assumed that the temperature should not increase by more than two degrees at all time. And we believe that the temperature rise induced by ultrasound, if the temperature rise increase of two degrees only would be in keeping with the thermal rise that is currently allowed for other biomedical devices. This is a way to tend to limit the thermal rise in order to keep it safe.
Another way to do so is to go back again to the diagnostic standards for ultrasound. For diagnostic ultrasound the thermal index was introduced. The thermal index is the ratio between the acoustic power that is emitted by the ultrasound probe, divided by the power required to raise the tissues target by one degrees. That is not so easy to use, but I will show you ways to estimate the thermal index.
First of all, there are different types of thermal indexes. One is the soft tissue thermal index, if there is no bone here on the pattern. One is bone at focus, and mostly this is when imaging babies, for example, in that case there could be some bone at focus. And there is a third case called the thermal index cranium or TIC, when the bone is close to the transducer. And this is actually the most relevant configuration for us for transcranial ultrasound stimulation.
And there are actually standards to calculate the thermal index. Here is an easier formula to estimate the thermal index, which assumes actually that all the energy that goes through the skull is entirely absorbed and converted into heat, and contributes to increased temperature.
So it’s a very conservative approach, but it gives you a very simple formula where this time W is the transducer output power, you can measure, you are supposed to know what is the output power of your transducer. The equivalent is the equivalent aperture at the skull surface, which is shown here, the interception of the skull surface.
Mostly for the diagnostic ultrasound the transducer is located at the surface of the skull. So usually D here in diagnostic ultrasound is actually the size of the transducer. In other case this is most of the time different, and the transducer is away, the equivalent diameter is this one which should be taken into account. And CTIC is a constant.
And I encourage you to carefully read the IEC 62359 standard to properly measure the TIC, and properly measure especially the equivalent diameter. But it gives you our data. And it’s possible to measure the TIC. And there have been several recommendations about the maximum TIC values for ultrasound images. Basically you can look at the values that were recommended by the American Institute of Ultrasound in Medicine, or the British Medical Ultrasound Society. They both came up with slightly different recommendations, and I’m going to summarize them in the next slide.
And actually in the next slide I will try to summarize all that I’ve discussed, in terms of going back to some existing standards, and trying to use them to adjust the safety levels. And this is work that has been done within the iTRUSST Society and the Safety Committee. So for mechanical index, what we believe is safe is to use a mechanical index lower than 1.9 in the brain and in the skin.
And the de-rating could be the 0.3 decibels per centimeter per millihertz recommended by the FDA, but we believe we could also go through other methods, and in particular we can take into account that it needs to be done in a conservative way, that you can take into account also the attenuation by this curve, to make sure that the mechanical index is less than 1.9 within the brain. This is for mechanical safety.
For thermal safety we can use any of those assessments. So I discussed that thermal rise should be less than two degrees at any time. It is also possible to cut thermal dose, and in that case, we believe it is safe to keep it much lower than the 7.5 limit to see a lesion, but to keep it below 0.25 CEM. And also here are the recommendations.
So actually once again it is a manner of power with the thermal index, and time. So it is possible to go actually as high as a TIC of five, but in that case one should not sonicate for more than 10 seconds, and if the thermal index is lower it allows you to sonicate for a longer period of time.
So hopefully this will be coming soon, we will provide consensus on biophysical safety for transcranial ultrasound stimulation. For that as I explained before, there are two ways to assess safety. One is this one, with consensus, which is a very practical approach to be able to start now applying ultrasound stimulation in humans in a safe way. And this is based on previous recommendations with other biomedical devices. It’s a very practical approach, but we are also aware that it is a very conservative approach.
What we’re going to discuss during the panel is that there is another way to do, and this was the first part of my talk, where I tried to look at the literature and look at the threshold in order to be able to induce either a mechanical or a thermal lesion. And as we discussed, the problem with this approach is that it depends a lot on the ultrasound parameters, on the frequency that is used, the precipitation frequency as well, and the total duration of the ultrasound.
So I would say this consensus has broad applications, it applies to almost all of the systems that are currently investigated for transcranial ultrasound stimulation. We need more thorough experiments in order to investigate all the possibilities. So I thank you for your attention.
ELIZABETH ANKUDOWICH: Thank you for that fascinating discussion. So we do have some time for our Q&A, and the question box is open for your questions. We have a couple that have come in during the talks. Jeff, you talked a lot about how some of these kinds of recommendations that we have that are informed by international standards and FDA kind of recommendations and guidelines are conservative. And there’s a question here that I think is interesting to discuss, how do we determine what intensity is considered harmful. And you’ve talked about how we need a lot more research to determine this. But how do we do this?
JEAN-FRANCOIS AUBRY: That almost needs to be discussed in the panel, but I can give you some -- First, be careful, intensity will not be enough. If we think in terms of intensity, it means that the only thing we could assess is the mechanical damage. Is if we increase too much the intensity, even with short impulses, you can induce some mechanical damage.
But if the intensity is low enough then you’re not going to induce mechanical damage. But even with a low intensity, if you sonicate for minutes, hours, then you might end up with thermal damage. So again, it is the combination of the two. I think we should first separate completely thermal safety and mechanical safety.
And then for both, the duration, and the peak negative pressure, so the intensity is important. So maybe an appropriate metric might be the total energy. And again, this would not be enough again, at least for mechanical damage. Even with very low energy we could induce mechanical damage.
The notice that we have is that at least we have lots of confidence in the mechanical and thermal safety for diagnostic ultrasound, because so far there has been no risk associated with ultrasound, at least without injection of microbubbles.
And so I agree, we could try to go higher in duration and higher in power. Now the question is what efficacy can we obtain with those very conservative parameters. And the thing is, do we really need to increase the parameters to increase the efficacy, as Kim mentioned, focusing through the skull is not easy.
So one way to increase the efficacy might be to ensure that we concentrate the ultrasound at the right place, and also that we compensate the aberrations in order to increase the peak negative pressure in the brain without increasing the energy on the transducer. So it’s also important I guess in terms of safety to make sure that we do not send too much energy just because we don’t want to correct for the aberrations. There are so many things to optimize before increasing the power or the duration.
ELIZABETH ANKUDOWICH: Thanks Jeff. We have another question for Kim. Given the neuromodulatory effects of ultrasound have been demonstrated in brain slices, in peripheral nerves in vivo and other isolated neural tissues, how significant is the auditory effect as a primary mechanism?
KIM BUTTS PAULY: I don’t know if I would say primary mechanism exactly. It is a significant effect in terms of sort of masking the effect of ultrasound for acute study. Ben Cop(ph.) and Lennart Verhagen did a very nice study where they were looking at for example whether or not ultrasound could have an effect on the motor evoked potential from transcranial ultrasound magnetic stimulation, so TMS.
And they found that it did have an effect on motor evoked potential, but so did playing a sound across the room had an effect on the motor evoked potential. So with these acute studies it seems like it’s very significant.
When you think about these longer-term delayed studies, even minutes or hours later, I think it’s a different story. I do think we need to be thinking about when we’re doing a study if we want to separate out the effect of the ultrasound from placebo effect, that then we’re going to have to think about it more carefully. But once we’ve done a study and we sorted out the effect from the placebo effect, the fact that they can hear it may not be an issue long-term for sort of delayed effects.
So with that being said, I do think we have to worry a little bit about auditory safety, and making sure that we aren’t playing sounds that are very loud, especially in an NMR environment where it’s coupled with something that’s already loud, and earplugs can make the sound much louder as well. So I think at some point we do have a safety issue even for that.
Could I just maybe, if I have answered that, comment on the other question that you asked Jean-Francois, That was back to safety. I wanted to mention the fact that we had done a study, there had been a study already in sheep saying there was a potential for microhemorrhages with neuromodulation.
So we did a study with sheep just to follow up specifically on that. And with our sheep, we were at pressures of three and a half megapascal, and we applied 8000 pulses. And we looked at H&E and we didn’t see anything, we didn’t see any effects of the ultrasound, we didn’t see any hemorrhage that wasn’t already there in the control animals. So that being said, there’s two things. One is what other studies do we need to look at.
And one of the things is that we have so many more tools now becoming available all the time. So we’re in fact going back and looking at the immunohistochemistry to see if there was anything there with glial cells or anything else for cytokine increases or anything else to be seen. In terms of the question of what do we need to do, we need to be very careful, and we need to be sure that we’re applying all the tools that we have. And I forgot my other train of thought, so I’ll just stop there.
ELIZABETH ANKUDOWICH: Thanks Kim. There is a basic question, mostly about some of the terms that were used. Are the basic parameters like frequency, intensity, et cetera, that are used in diagnostic ultrasound, different from those used in therapeutic ultrasound for neuromodulation?
JEAN-FRANCOIS AUBRY: Not much actually. It depends on the studies. Some differ, some are within the range of the diagnostic ultrasound parameters. The main difference though is that for diagnostic ultrasound the impulses are very short, one microsecond.
So of course, there is a reason why for decades people have been using ultrasound to image the brain, through the temple mostly, and they never induce neuromodulation, because they were using low intensities and very short sonications. Even when using power Doppler for example when they use more sonications and a bit longer signals, they didn’t.
So if you restrict completely to I would say the parameter space that is allowed for diagnostic ultrasound, there is little chance that you will induce neuromodulation. But if you respect the mechanical index for diagnostic ultrasound, and if you respect also the thermal index, then even though you are under the safety limits of diagnostic ultrasound, it is possible to insure a neuromodulation effect.
KIM BUTTS PAULY: So the question about are we using the same terminology, we are trying very hard to use the same terminology as with diagnostic ultrasound. And I think there is only one case where the terminology has been a bit confusing. I think most of the terminology is all exactly the same. The one thing I think is a bit confusing is the use of the term burst. In diagnostic ultrasound it refers to a single pulse, and then people in transcranial ultrasound stimulation say a train of pulses is equal to a burst.
So we’re actually coming out with a paper on standardized reporting, where we’re suggesting people don’t use the term burst, because it has been very confusing. And then just kind of reiterating what the other terms are. And I think Jean-Francois maybe you can comment as well if there are places we have deviated from diagnostic ultrasound.
JEAN-FRANCOIS AUBRY: I don’t think so. We tried not to deviate, so the terms are the same.
ELIZABETH ANKUDOWICH: Kim, would you say that MRI-ARFI could be used with certainty to confirm delivery of transcranial focused ultrasound in the human brain?
KIM BUTTS PAULY: With our sheep study it worked beautifully, because there was once or twice when the equipment just wasn’t working, we didn’t know that until we were like we don’t see an ARFI spot, and then we figured out it wasn’t working. So it was absolutely just a very robust confirmation that it was working, and we got a nice focus, and that was beautiful.
So I do think that it can very much be used kind of as a confirmation that way. The thing that is still going to be tricky and needs a little bit more research is trying to back out what the intensity is from the MR-ARFI spot, because it’s just a little complicated depending on the stiffness of the tissue and how big the focal spot is and how deep, and exactly how much it displaces. That being said, I am very optimistic.
ELIZABETH ANKUDOWICH: There is a lot of interest in terms of questions here of safety, of ultrasound, on the developing brain. And I’m going to kind of put a bunch of questions together. Have they done any estimated pressure models in children or adolescent skulls, or have you just looked at this Kim in adult skulls? And then also, are there unique biophysical safety concerns in using these types of tools in young children or even infants? Any ideas here?
KIM BUTTS PAULY: I am going to say I don’t know. We haven’t studied developing brain. I think that I am going to defer to some of the clinicians maybe. But I’ll just fall back on the fact that we have used diagnostic ultrasound on fetuses for millions and millions of studies.
JEAN-FRANCOIS AUBRY: Anyway, it will probably take some time, before we do apply transcranial ultrasound stimulation to infants, and even to people under the age of 18. Especially, I don’t know about the US, but in France, until the proof of concept and safety and efficacy has been assessed on adults, we don’t use those on young people.
ELIZABETH ANKUDOWICH: I think that makes a lot of sense. I’m mostly curious about what people are thinking in this space. A final question, and then I will turn it over to you, Jeff, for the panel discussion, is are there current recommendations for how best to monitor potential thermal damage and mechanical damage in humans, other than choosing stimulation parameters below the recommended safety limits?
JEAN-FRANCOIS AUBRY: I would leave this question to the panel discussion. That’s exactly what needs to be discussed all together.
ELIZABETH ANKUDOWICH: That sounds like a great segue actually. Jeff, would you like to go ahead and launch the panel discussion?
JEAN-FRANCOIS AUBRY: Sure, with pleasure. SO I would like to ask all the panelists to turn their mic on and video on. Thank you. So feel free to jump if there is any question you would like to answer. Maybe altogether, if we initiate the discussion with the Insightec procedure. When using the (inaudible) neural system, the procedure consists, this is for thermal therapies, they increase slowly the temperature. They don’t jump directly to a high temperature. And they see transient effects on neuromodulation.
I think it is interesting to discuss this, because there is, clearly in that case we have thousands of patients that have been treated, for which it started with a transient effect, and then they went all the way to a permanent effect, inducing a lesion, and actually seeing a permanent lesion. Kim, you attended a lot of treatments, could you comment on that?
KIM BUTTS PAULY: One of the things that was really striking about that is that as you raise the temperature slowly, because initially what they want to do is get it to about 45-48, make sure it’s in the right spot and it’s nicely localized, and then raise it up to around 50, and it starts having reversible symptom changes.
So for example in that case, for the central tremor, if the subject says well I have tingling in the lips, then the neurosurgeon says oh I know I am one millimeter too far posterior and I need to move it anterior one millimeter. So it’s super to have those reversible symptom changes. And that’s because there’s a small temperature rise.
So if we think back when I was talking about how the lipid membrane is sensing a mechanical force, but it’s also that those order domains are sensitive to temperature. I think it’s entirely possible that it’s the temperature that is having an effect there, and that what we’re seeing with those treatments and those kind of reversible symptom changes is exactly what we’re potentially looking at. And your data Jean-Francois might also support that.
JEAN-FRANCOIS AUBRY: We used actually the Insittech machine with accelerometers attached to the patient’s hands. And just before inducing permanent lesion, what we could see is that increasing the temperature to 42 degrees, the tremor was decreased significantly, but it lasted only for a few minutes, and then when we were coming back to the next sonication the tremor was back. When increasing about 56 degrees is when we would see actually a permanent decrease in tremor.
And Kim, what you mentioned before doing so, instead of using continuous sonications we also tried burst sonication. And in that case we showed evidence that it was possible to reduce the tremor drastically. We saw a 98 percent tremor reduction. And this is actually, we think the space of parameters that I mention in my recommendations, with low peak negative pressure and also no thermal rise. The peak negative pressure was lower than one millipascal. And the thermal rise, we were using actually MR sequences, we just could not measure the thermal rise, it was below the MR noise.
So this gives some evidence that it’s possible to induce a drastic effect on a patient without trying to increase the pressure too much, saying within the recommendations for MI for example, current recommendations for MI, and with very short sonications for which we could not even see a thermal rise.
So the only thing is we see the difference between burst sonication and continuous sonications. It was surprising to us that again the (inaudible) system has been used on thousands of patients, and a reduction in tremor was never seen, not as efficiently as what we saw. And what we obtained was with a few watts, very low wattage, and short duration.
So I was mentioning the Insittech procedure because those are also valuable data that could be used in order to assess the threshold for lesioning in the brain, because we have all those data for which we have actually the thermal rise that was measured with the system. And we could compute for example the thermal dose and see at which level we start to see a permanent effect. Talking of which, one other question is how to assess tissue damage, or tissue effects. And should we look at H&E for example. Could anyone comment on how, what are we supposed to see, and what could be the marker for tissue damage? Maybe Matthew.
MATTHEW MYERS: I think I would ask a slightly different question. I would say, instead of tissue damage, let’s ask when do adverse bioeffects occur. Because what we think is going to happen, as we move away from these conservative metrics for safety, and people increase the mechanical index for example, is that we might see adverse bioeffects before you see traditional measures of damage, like cooked cells or cells torn apart by biomechanical damage. And I think the adverse bioeffects will manifest, well in animal models they will manifest in terms of behavior.
So if we’re going to do animal models then I think behavioral measures have to be an important part, because you can get adverse behaviors before you get traditional mechanical and thermal damage. And I think there are electrophysiological measures too that start to manifest before you get, again, thermal damage and mechanical damage.
And I think you can see kind of subtle H&E measures of inflammation before again you’re seeing gross damage. So what I would advocate is that we ask what are the adverse bioeffects where the nervous system just isn’t functioning anymore, but it’s not being torn apart either.
JEAN-FRANCOIS AUBRY: And how do you make the difference between, with the fact that the nervous system is not functioning anymore, or is just functioning differently, because that’s what we want, we want to induce a change. And sometimes in some experiments people are trying to induce movements, to induce changes in behavior.
MATTHEW MYERS: Yes, you want to induce some kinds, and you don’t want to induce others. There are definitely behavioral measures in mice for example that are undesirable. There are coordination measures and just measures of how the animal is sort of taking care of itself. I think there are definitely adverse event metrics that one could use.
JEAN-FRANCOIS AUBRY: There could also be some more subtle changes I would say on tissue and maybe on the immune response. Elisa, maybe you could say a word.
ELISA KONOFAGIU: Along those lines, we have done a lot of work on the brain, and when it comes to damage we have looked a lot at the blood-brain barrier, which I know was not mentioned today, but it does have modulatory behavior changes in the brain. And we have seen that not only in chemical or immune response activation, but also in behavior. And along the lines of what Matt was mentioning, we have tried to see whether it was safe, number one.
And then once we identified the safety range, which was basically H&E completely clean, so that means no red blood cell (inaudible) in short term or long term. So that means within 24 hours, which is usually the highest, the peak of the damage, recovery hasn’t started yet, or dark stained neurons, which typically is more permanent.
And we have seen the opposite, when you operate within that safety range you can actually change for the better working memory in mice. So you can enhance, not too much, but you can definitely enhance it. And then also some gait attention, and gait changes in nonhuman primates. Again, very subtle, but it’s there.
So there is an enhancement that is happening as a result of both neuromodulation, so very little opening and opening that comes from microglia and microphages, and we’re trying to understand how they’re mechanically sensitive. So these cells are mechanically moving in our brains, and they’re basically constantly detecting whether there’s any debris or in Alzheimer’s plaques or tau or any other toxic proteins that could compromise the brain behavior.
We have seen that they can be mechanically activated by ultrasound. So in a good way. Obviously when you’re increasing density and all that good stuff that we just talked about, you can also damage and cause proinflammatory responses, which we don’t want. But there is a window, and that’s very important to note it’s a very short window but still there that we can potentially also manipulate.
JEAN-FRANCOIS AUBRY: I mentioned in my talk, I tried to find some work showing some threshold for tissue damage. It is true we also have lots of data that has shown safety. So maybe we would need kind of a catalogue of results showing actually not only, not focusing on the negative aspect, either lesions or behavioral change, but also what has been reported so far as safe parameters, to kind of try to narrow and determine exactly this window, we want it to be efficient and safe, and maybe look too much on the upper range of lesioning, it might be also limiting in terms of determining this window.
ELISA KONOFAGOU: It’s definitely not an on-off switch. So basically, you can go from absolutely safe and nothing happens in the brain, all the way to of course what is FDA approved, which is ablation. But in between there is a granular aspect which is very interesting, which you can dial up either the pressure, the intensity, the duty cycle, or the pulse duration, the PRF, and then enter in different regions of where you want to be, they’re neuromodulatory or immunomodulatory, or sometimes damage, but it can be reversible because the brain is resilient and as we know from concussions can mend a lot, and also from electrode placement, when you do DBS for example, measurements or therapies, you are damaging a lot of the brain just from the insertion of these electrodes. But then the brain has a way to basically repair. So that’s very interesting from that one if you don’t want to have this binary way of treating ultrasound effects on the brain.
MARK SCHAFER: I just want to comment, we did some work with patients who are having surgical resection of epilepsy centers in the brain, where we exposed the tissue to different levels, and then we were able to resect the tissue and do H&E thereafter. So this was really in human trials. A small number of patients. But we were able to go up to at least four times the FDA guidance without any indication of any damage on H&E stains.
And I did see something in the chat about inducing seizures. I believe as I recall that study, although seizure control was not one of the endpoints, there was I think one patient who did trip into a seizure. But there were also some other patients who found some relief in the day between their ultrasound treatment and their surgical resection.
So there is this window, as Elisa and others and Jean-Francois have commented, between the FDA guidance and ablation it is granular, and I think it is a perfect way to put it Elisa, there are different ways to approach those locations in multi-dimensional space and PRF and intensity.
But I do want to point out, everybody calls the FDA limits safety limits. They are regulatory limits. They were not established on the basis of safety, they were strictly regulatory. Now, they have proven to be very safe. Billions of ultrasound scans over the last 50 years have shown an incredible safety record, unlike almost any other medical technology. So operating there below is almost a guarantee of safety, and then it’s just how far above we can go. I do believe there is room to move above that. But we should of course do it carefully.
We know there’s a top end, at ablation, and you’re quite safe here below. I think we can creep up reasonably safely without getting too wild about it, based on the data we’ve got so far. And I think Jean-Francois’s presentation about Mis are very significant in order to get any kind of in a non-microbubble environment, microbubble MBD is a whole different category. But I think we have plenty reasonable room to operate.
JEAN-FRANCOIS AUBRY: If I may, we have some room to operate, but not that much. I mean, it depends on the duration of the ultrasound. For one second duration, the MI, I know it’s difficult to talk about MI in that case, but it was something like four. So it’s not so high compared to the 1.9 limit. When we discussed it all together within the iTRUSST consortium to decide to kind of validate the MI for longer duration, so we are already pushing the limits when we say so.
MARK SCHAFER: And also recall the MI as calculated, you brought it up, it’s a very good point, the FDA derating is in fact a completely arbitrary derating value, which I believe a different derating we talked about in that paper that is more appropriate for the brain case with the skull different attenuation in the brain, and the paper you referred to of course was just without the skull.
So one of the problems in this whole field is we’ve got a lot of apples and oranges out there in terms of experimental data and what we’re trying to do versus what was done before. I think we’re treating it very seriously and walking very slowly, as we should.
But I believe we will get to closer, I think a lot of the variations in efficacy we see now are because we’re near thresholds of efficacy, because we’re operating so well down in the exposure category. I think we need to just get a little bit higher, and I think our efficacy will go up faster than our safety will go down, in my opinion.
XIAORUI TANG: I have something to add in addition to thermal and the mechanical damage as Matt just mentioned, I’m with FDA, we also pay close attention to the adverse bioeffect. When we’re talking about treatment, ultrasound must be applied to patient population. So we need to pay close attention on the safeties of ultrasound on different patient population, for example when we apply to seizure population, epilepsy population, or major depressive patient population. So these people may have different parameter tolerance than the other patient population or healthy patient population, I would just like to add on that. Thank you.
MARK SCHAFER: Certainly patients subject to seizure would be a special category. But there are things that we would do in a different population that would then cause an adverse effect and then be the cause of their underlying condition, certainly. And I forgot to identify myself, I’m with Drexel University. Sorry.
ELISA KONOFAGOU: I also want to add the age factor, because we’re looking at mice that are more aged, and mice that are younger, and the skull tends to be more homogeneous, and even if it’s tiny and thin it does make a difference for both the attenuation that goes through and the imaging aspect. So it’s important when we work on preclinical models to not only work with young mice but also mice that have either compromised brains as a result of disease or aging, but also the skull aspects.
JEAN-FRANCOIS AUBRY: And it’s even worse considering a human population. As Kim mentioned, from patient to patient the skull will vary a lot, and so will vary also the transmission of the ultrasound. So the dose that we apply to the brain which will affect both the efficacy and the safety of the treatment.
But also, as Kim mentioned, the defocusing effects. For some patients we might actually target the right location, for some we might be off. And it might also have also implication again on the efficacy but also on the safety, depending on which part of the brain we’re currently stimulating, it’s important to make sure that stimulation will occur at the right place. If we discuss safety this should be included in the whole package.
MARK SCHAFER: Certainly work was done in many years of trans-cranial doppler where the patient studies have shown ethnic variations, age variations, sex variations that have quite dramatic differences in the bone characteristics. So Jean-Francois, you’re exactly right, and that’s something we should be aware of and recording the data as we do these studies going forward.
KIM BUTTS PAULY: Could I just ask one of the panelists, I’m curious, Matt, if we ask you, over and above the things that we’ve already talked about, are there other things that FDA would like to see?
MATTHEW MYERS: I think that Mark brought up a good point, that a lot of FDA’s wisdom out there, guidance if you will, is arbitrary. It was based on, well the diagnostic guidelines or just regulatory guidelines, they weren’t based on science. So what we would like is a real science-based approach to determining safety for neuromodulation.
And unfortunately, that’s going to require a lot of studies, because there are so many relevant parameters that we’re seeing. The parameter space is probably seven or eight deep. But I think it needs to be done so we can sort of leave the diagnostic guidelines behind and have real science-based measures of the safety envelope, and the efficacy envelope. So that’s what we’re looking for right now.
JEAN-FRANCOIS AUBRY: Who should conduct those experiments? This is an extraordinary amount of experiments if you wish to span the whole field, in terms of frequency duration, it’s a lot. And so far, some of it has been done in terms of efficacy. But I remember a long time ago I wanted to do it, and actually Kim did it, but just adjusting the duration and the frequency is huge.
If you add on top of this the fact that assessing the efficacy is one thing, but the safety, so many things can happen and so many things need to be monitored in terms of H&E, damage to the brain, but as you said also behavior. So it is a lot of work for one single set of parameters, in terms of if we said just duration and frequency. So how could someone collect enough data to cover all the frequencies and other parameters that are currently being used?
MATTHEW MYERS: It couldn’t be done by one study, to be sure. It would have to be a very group effort. But I would say the alternative is that when people come to the FDA and we don’t have a good grasp of the safety envelope, what do we do, we require them to do animal studies for the parameters that they are using in their study. And that’s not very attractive either, that’s burdensome.
So yes, it’s a big investment. Hopefully we have funding people in the audience now. I think it’s a very worthwhile audience, I mean it’s worthwhile to the audience to invest that money, so that our knowledge base is better and our regulatory processes is faster, and we’re not just taking submissions on a case-by-case basis and saying all right, do the animal studies with the parameters that you’re proposing.
So I agree, it is a very large study and a big investment. I think that modeling can help break it down. I think if we had a better mechanistic understanding of neuromodulation that would help us to say we don’t have to cover the whole space, we think we understand the dependence on different parameters from models. So I think we don’t have to really cover the whole space, but I agree it is a big investment.
MARK SCHAFER: The effort that went into developing the current thermal index and mechanical index was a consortium of almost all the diagnostic manufacturers, the AIUM, and the FDA, and the university community working for two and a half years straight through. So just getting to that level, and at the end of the day being stuck with the 1976 output levels. So this would be a major undertaking. But certainly, worthwhile if it gets us closer to some answers.
JEAN-FRANCOIS AUBRY: But in a sense, for diagnostic ultrasound, it was maybe simpler, because what was considered was considered only short pulses. In our case pulses can be longer. And then there is also the pulse repetition frequency, and the total duration of the sonication. So the space of parameters is much larger for TUS than diagnostic ultrasound, when trying to test on (inaudible) for example, to assess the safety.
And it was, as you mentioned Mark, it was a huge task. It is very impressive, what has been done on diagnostic ultrasound to kind of try to assess the safety. We are now at another level, I would say not only in terms of pressure, but in terms of I would say variability of parameters that are currently used.
MARK SCHAFER: I agree. We made a decision early on to separate the field, no pun intended, into short pulse, look at mechanical effects, and then longer pulses. Because we’re dealing with doppler, we’re dealing with continuous wave doppler and pulsed wave doppler on single beams focused at a single spot for a long time.
We said that would be on the thermal side, and short pulses for the mechanical index, and we did divide that. Now we’ve got a situation where pulses of not a dissimilar intensity or peak pressure from the imaging side are now being used as longer pulse ensembles. So indeed, bringing the two and meeting somewhere in the middle. It is a different space, but some of the basic physics is still there that underlie everything.
KIM BUTTS PAULY: May I ask, Jean-Francois, the panelists a question? And I know we are going to have another session where we talk a lot about seizure, we are going to get some experts from neurology for example. But maybe we could just ask this group, is there, especially Mark and Matt, is there any data that has already been shown about diagnostic ultrasound in terms of inducing seizure? Is there any data there on ultrasound induced seizure.
MARK SCHAFER: Except for the study we did where we were excising the tissue, I would have to go back and look at the raw data. I think we had one subject out of eight that the seizure occurred. And then we had two subjects for whom their rate of seizures went down afterwards.
But I don’t believe that diagnostic ultrasound, again the whole idea of imaging the brain is relatively new in diagnostic ultrasound. Transcranial doppler has been around for a while. I’m not aware of any case reports of seizure. And I’m also not sure whether at the time in fact epileptic subjects, that was considered exclusion criteria for transcranial doppler, to tell you the absolute truth. So I’m not aware of anything, but it may be a fluke of the way the products were cleared.
JEAN-FRANCOIS AUBRY: This will be discussed anyway in the next panel, and we are now running out of time. I think it is time to end this panel. I thank you a lot. I thank all the panelists for their inputs. And we will now leave it to Lizzy, we’re going to have a short break I think. Thank you.
Session 3: Physiological and Clinical Considerations
ELIZABETH ANKUDOWICH: Hello everyone. Welcome back to our third session, focused on physiological and clinical considerations. Our first speaker in this session is Dr. Shy Shoham who joins us from the New York University Grossman School of Medicine, and who will be speaking today about biomechanisms.
Next we will have Dr. Leonard Verhagen from the Donners Institute for Brain Cognition and Behavior at Radboud University. Dr. Verhagen will be speaking on physiological safety and will also be our moderator for this session.
He’ll be followed by Dr. Fidel Vila-Rodriguez from the University of British Columbia speaking on clinical safety and mental health therapy. Welcome, Shy. Thank you for joining us today. And feel free to share your slides.
SHY SHOHAM: Thank you very much. I am going to be talking today about FUN Biomechanisms. Fundamentally what we would like to know is the relationship between the ultrasound stimulus and the evoked effects. So, different than the safety. We’ve talked a lot about safety. Let’s talk about bioeffects here. And most earlier studies in this field basically looked at measured of evoked activity that were aggregated or downstream readout activities, such as behavioral responses, EEG and FMRI functional signatures.
Now, to do this mapping is something that would help us to optimize protocols for example, or design better experiments. And as an engineer I think this is something that is very basic and near to my heart, because essentially in every engineering discipline nowadays we use these types of design software that basically use such a mapping.
But somewhere between the ultrasound stimulus and the effect we have some sort of a biophysical effector, maybe part of a cell or the membrane. And then we also have the interaction with the neural circuits. So we would like to capture all of these things to really understand fully the mapping.
It turns out that this turns out to be rather challenging for various reasons. The first set of reasons has to do with, and Elisa Konofagou talked about it, is the high dimensionality of the parameter space. We have frequency, amplitude, envelopes. We have parameters like pulse repetition frequencies, the duty cycle, the duration, and depending on how we build the pulse trains there could be additional ones. These different parameters all have effects.
And every study that characterized these effects is shown one or several types of parameter dependence. And in fact, the space of parameter dependences is not simple. For one thing we know that some parameters elicit activation, some of them elicit suppression. And potentially the parameter space is way more complex than what I’m showing here.
A second set of challenges has to do with the physics and physiology, and Kim has alluded to that partially as well. First of all, not just in safety-wise, but also effect-wise, we always have mechanical and thermal effects that are elicited jointly, and it turns out that even relatively subtle thermal effects of one or two degrees Celsius in some brain regions actually block neural activity, there are several lines of evidence on that. In addition, the ultrasound fields are relatively complex, and we have multiple types of mechanical effects, pressure, strain, radiation force, cavitation, and Kim has expanded on that.
It turns out that almost all of the basic physiology toolbox that has been used successfully by neuroscientists to characterize a lot of other phenomena turned out to be fairly challenging when applied to ultrasonic neuromodulation.
For example, tools for doing intracellular electrophysiology, glass electrodes tend to resonate, in vitro setups have multiple reflections in them. And then of course auditory confounds have been really highlighted when it comes to small animal in vitro as well as human neurostimulation. So just the physiology toolbox turned out to be challenging.
And then as a result of some of these things as well as additional challenges, we still have an open question about what is really that target effector. And I just want to highlight the possibility that it is a complex collection of calcium channels or other ion channels that operate in concert. It could be in neurons, it could be in astrocytes. And those are extremely small, and difficult, and also differentially distributed in different neural tissues.
So this is the set of challenges. So if we circle back to this reverse engineering problem, what we see is we have a multidimensional parameter space, multiple mechanical bioeffects, interweaved with thermal bioeffects, affecting networks but also having significant confounds, and then that leads to our measures, which are diverse and sometimes have their own challenges.
So I started off with a disclaimer, but we actually can do things to characterize these bioeffects fairly carefully. For example, in recent work jointly with Daniel Razansky and Hector Estrada, we’ve built a high resolution phased array-based system for imaging with wide-field fluorescence imaging, for imaging ultrasonic targeting. This is in the relatively high frequency, so thermal effects actually dominate, and we can see here a thermal dip that is elicited by very short ultrasound stimulus.
But we can also simultaneously explore ultrasonic neuromodulation that is evoked locally by this focus. So we have one kind of tool for studying evoked neuromodulation, but what actually happens at the network level in this localized excitation spots? To address this question, we combined focused ultrasound with two-photon imaging.
And basically, in transgenic mice that express GCAM fluorescent reporter in specific neurons. We surgically implanted the cranial window in the cortex and coupled a microscope platform with a ring-based ultrasound transducer where we could image through this ring. And we acquired fluorescence videos of the functional responses of visual cortical neurons during ultrasonic neuromodulation.
This is what a response looks like in a line that expresses GCAM success in SST neurons. We developed an analysis pipeline to analyze these movies, and basically could see very robustly evoked responses in a population of neurons when we delivered these ultrasound stimulus. We then first addressed an obvious concern, which is the question of the focality of the evoked effects.
And what we did was first we mapped the size of our evoked ultrasound field, it’s laterally about half a millimeter, axially about three millimeters, directed as I said at the mouse visual cortex, and then we could displace our transducer laterally in different directions, and characterized how that affects the responses, and we saw a dramatic decrease of the responses, even when we moved by as little has half a millimeter, suggesting it is a very focal response with some network crosstalk laterally.
After verifying that we could elicit robust and focal effects, we started comparing the responses between the different cell types. And we specifically looked at three classes of cortical neurons, the Thy1 pyramidal neurons constitute the majority of the population of neurons in the brain, and these are excitatory neurons. Somatostatin inhibitory neurons that target excitatory dendrites, and peripheral human inhibitory neurons are identified as fast spiking neurons, and have generally the faster spike waveforms that were studied by different groups in this field.
So together these three cell types constitute more than 90 percent of the total cortical populations, and considering both excitatory and inhibitory neurons, I’m going to highlight the behavior of the Thy1 and SST populations in what follows. We looked at the effects of different parameters.
The first parameter that we looked at was pressure, so intensity. And we saw a monotonic increase in the response magnitude for increase in pressure levels. This looks like a roughly quadratic dependence in the regime that we explored. We then looked at duty cycle. And again in both populations we see a monotonic increase in response magnitude from increasing duty cycle. But this time we actually brought it to saturation and at high duty cycles. So this fits more with a sigmoid profile.
We wanted to know whether these monotonic behaviors can be summarized with some sort of a unifying scaling law that captures growth of the compound effect. So our pressure and duty cycle biometric suites, while seemingly independent, actually both of them jointly map to a common measure of acoustic dose. For example, the spatial peak time average of the intensity is plotted here for the two different parameter suites.
So we wanted to know which measure of acoustic dose is best suited to describe the neuromodulation effect. To answer this question we can take our two response curves, obtained from each parameter sweep, and use a function to project both curves into the same dose space. If dose measure X accurately captures the compound effect of pressure and duty cycle, then we expect to see a very similar functional relation. If this dose measure does not we expect to see a divergence. And this is plotted here, these two possibilities, F1 and F2. This was our recipe.
And we applied different measures to our dataset. And what we found is for example if we look at the average pressure, there is some discrepancy, first looking at SST neurons, there’s a much lower discrepancy when we look at the temporal average of intensity.
And again, there is an opposite type of discrepancy when we look at the RNS intensity. So, for three different compound measures we see different levels of complexity. We saw a very similar picture when we looked at the different cell population of the Thy1, the pyramidal neurons. So this suggested to us that the average intensity, the ISPTA, the time average intensity, the pulse time average intensity, is a great compound measure.
Then we started looking at what would be the net effect with a given ISPTA. So our different cell populations, the excitatory Thy1 and inhibitory SST had different response magnitudes. They also had different population density, there’s more pyramidal neurons. When we factored those two things in, we got sort of a tug of war between excitation and inhibition.
And that led to a net effect curve that had net suppression at low ISPTA, it’s also low duty cycle in this case, whereas net excitation kicks in at higher intensities and duty cycles. So this is something that comes out of this single cell level recordings in mouse, and you might wonder if it also says something about human ultrasonic neuromodulation.
So here I defer to studies by Winn Legon that combined TMS, motor evoked potential by TMS, with focused ultrasound delivered here in the center. And what Legon and colleagues showed is a very similar curved dependence for the effect of ultrasonic neuromodulation on these motor evoked potentials, with the suppression kicking in at low duty cycles, and excitation kicking in at high duty cycles. So at least on the face of it there is an agreement.
These results are also interesting to examine in the context of the first question which I posed, which is the site of the effector somewhere in the neuron, and the mechanism of action, which of the mechanical effects of ultrasound actually induces this neuromodulation. There were several key ideas in the literature. For example, the effect could be mediated directly on ion channels, or through the membrane indirectly in ion channels, but still the effect in both cases is ion channels leading to depolarization or calcium entry.
There are biophysical ideas on how conformational changes in the membrane can induce changes of polarity. These are known as piezoelectricity or flexoelectricity. And then another idea that my colleagues and I contributed to the literature is this idea of intramembrane cavitation, where ultrasound induces bilayer cavitation to trigger membrane capacitive currents.
A little bit of the history of this last idea. This was a brainchild of my colleague Eitan Kimmel, who basically recognized that by plugging in at the relevant physics, biophysics, that the membrane could potentially be the element capable of absorbing ultrasonic mechanical energy a little bit like how the windows of this building would vibrate if an airplane nearby would produce a hypersonic shock.
We embedded this idea together with a very clear kind of a biophysical reality that we have a lot of membrane proteins anchoring this lipid membrane along islands or other kind of structures. And basically, posited that during ultrasound induced vibrations small structures in the lipid membrane could be potentially vibrating. Of course if that vibration is taken to an extreme limit it could lead to tears. But that is not the regime that we are working in. And indeed, observations in tissues that were sonicated suggested to us that this could be happening in sonicated tissues.
When we examined what that would do in terms of the biophysics, we realized that these high frequency membrane oscillations, when coupled electromechanically, would lead to progressive depolarizations and a neural response. This was coupled into a Hodgkins-Huxley type of model, and what we saw was basically a slow accumulation of charge on the membrane, ultimately leading to firing of action potentials.
We tried to examine predictively what would come out of such a mechanism. And this idea of progressive charge accumulation, you can intuit it as a sort of Archimedes screw that rotates and rotates and basically gradually leads to the elevation of water across a threshold. So we started examining what would be predicted by such a mechanism.
And at that same time that we were playing with our model, the group of Kimm Butts Pauly performed the first parametric study of ultrasound induced in vivo effects, leading to interesting parametric curves on the dependence of intensity and frequency, and our model predictions were very similar to that. In fact when we corrected it from some very simple corrections, it really fit very well these parametric behaviors, both in terms of dependence on intensity and on duration.
We then examined what would happen when we applied it to different cell types. And different cells have different cocktails of membrane proteins that define their biophysics. For example, cells that are known in the computational neurosciences as low threshold spiking neurons, or LTS neurons, also identified mostly with SST inhibitory populations, have prominent expression of T-type calcium channels, which change the biophysics, especially it changes the impact of short pulses on these neurons. So then when we examined the effect of different duty cycles we saw that all key populations of cortical neurons would have low thresholds at high duty cycle, again in this putative model.
But when you go to low duty cycle, to sparse relation, the threshold will be low only for this low threshold spiking neuron. So they will be basically the ones that kick in at low duty cycle. And this was the first kind of prediction of cell type selective activation. It also explained a gamut of results from the literature that basically showed inhibition or suppression dominating at low duty cycles in this complex parameter space.
And of course, this circles back to what we saw before, I don’t know if this effector model is correct, but it is at least consistent with our findings on the differential acquisition of the SST neurons versus the pyramidal neurons.
So to summarize what I talked about, I presented our platform that allows dense optical inspection of FUS-evoked effects in vivo. The response seems to scale well with the temporal average of intensity. And I plan to provide bidirectional control over the net circuit activity with cortical neurons. Predictive modeling, in our case it was based on the NICE model, but I think more broadly other frameworks are also possible here, could add crucial insights to our understanding of the bioeffects.
And then of course I want to highlight that most information is still missing. For example, we still don’t know for sure what are the key biological effectors, we don’t know what will be the difference between rodents and humans. Different brain regions behave very differently, and Keith Murphy showed us a beautiful example of that, examples of that. And then for many of these, since we’re in a mental health conference, most of the effects of interest are these offline effects, and we actually know very little on that. And there are of course many unknowns.
I would like to thank the people who helped me, that really did this work. The two-photon work was done as a group effort by Yi Yuan, Theo Lemaire, Justin Little, Amy Lemessurier and Rob Froemke in our group. And I also showed interesting insights from the work of Dan Razansky and Eitan Kimmel and colleagues.
ELIZABETH ANKUDOWICH: Go ahead, Lennart.
LENNART VERHAGEN: I am Lennart Verhagen from the Radboud University and Donders Institute. I am going to talk about the physiological safety of transcranial ultrasound. But first, these are some of my funders, and there is no financial conflict of interest. When I’m saying physiological safety, we’re really extending and moving beyond the previous session.
Earlier we were talking about structural damage and lasting impairment. We already learned that there is a very large transition period. That’s something that we’re going to dive into now, the secondary effects that might lead to adverse events, and the risk of the transient effects.
This talk will start with some boring definitions I am afraid just to get all of us on the same page. Imagine that we have an intervention. The effect of your intervention could be ultrasound stimulation, it’s dependent on the stimulation parameters, it’s also dependent on your patient or participant, the state that they are in at the moment, or the context of your stimulation.
And in the end what you’re looking for is what we call a primary effect, the intended reaction of your ultrasound stimulation. And maybe you’re measuring that in physiology, behavior symptoms or clinical outcomes. The primary effect can be found at many different levels. But we have learned a lot about the biomechanisms of these primary effects, about the neuromodulatory potential of ultrasound, in Shy Shoham’s talk just before.
Now we’re going to look at whatever else is left for secondary effects. And in some cases they are going to be driven by the neuromodulation of ultrasound, they’re going to be related. Other secondary effects can be unrelated, and then of course we have suspected or unsuspected ones. Now, when these effects are related we call them a secondary reaction. And we can scale them on anything that could be happening.
But of particular note are those adverse events. Then the secondary event is also untoward and it is a medical occurrence, it has medical relevance here. And yet as you can see, adverse events can both be reactions and unrelated. The reactions we call adverse reactions, of highest interest are serious adverse events, where this leads to either death, deformation, lifelong changes, hospitalization.
So the range between related and unrelated, as you can see here at the bottom, that already is a large scale, and I want to emphasize that maybe we can separate it further out because we can try to investigate if effects are related to the stimulation itself or the participation of your study or receiving or being a participant here at all.
A second domain in how we may classify the secondary event is whether they are expected or unexpected. I am mostly going to talk about expected secondary events here. And maybe even the panel discussion will be coming to what we might be missing here in the overview.
So let’s blow this up, and let’s start filling this with reported evidence. Ultrasound is a novel application of a noninvasive brain stimulation in humans, and it is particularly helpful for example that there a single center retrospective study of secondary effects out. This is a questionnaire list. There is no distinction between sham or real separation, but subjects are asked to report on all of these questions, and later on whether they felt it was related or not.
And you can see that participating in an ultrasound study is particularly boring, it leads to mild sleepiness as the most common secondary effect. The dark blue here means this isn’t reported. A total of 65 participants, seven studies in a single center. I also want to report other reported studies, but here now a systematic review of human applications of transcranial ultrasound. This review was mostly focused on mapping out the primary effects. That’s actually what the color coding is, whether it’s a clinical, neurophysiological change.
So you can see the whole solar burst plot of primary effects, but they also were counting secondary effects, and including adverse ones. So let’s put them down. Most of them are actually related to the participation. It is boring and somewhat tiresome. In these studies, fatigue and physical constrains are most often reported. And this can lead to headache or muscle ache, or attentional changes, mood deterioration, dizziness, even anxiety and sleepiness.
I want to highlight that in all of these reports, actually we are unable to tell if they are direct effects of the ultrasound stimulation. This really necessitates that we compare good control conditions against the interventions.
So sham or even better, a representative closely matched with peripheral effects matched as well, so that the subject can be truly double blinded. That is often not the case in currently reported studies. That might make us a little bit sad that we do not have that information yet, but there is something, namely, we can also see what hasn’t been reported.
So 704 participants across these 30-something odd studies, and there are no severe or persistent adverse effects reported in any of them. That doesn’t mean that everything is fine, there is nothing to worry about it. In fact this is a call to all of us researchers in transcranial ultrasound to very carefully always report our secondary effects, especially our adverse events, and start mapping out where we’re delivering what kind of ultrasound, the exposure. Because in the end, we would like to start quantifying on- and off-target effects. I’m just giving one example.
There are quite a number of studies in humans, our researchers are focusing the ultrasound on the amygdala, and this is a simulation that you can see over here. You might also wonder what, if we would focus too deep, how much energy would then be deposited in the brainstem, what are the secondary effects of these off-target stimulation.
And a lot of this will be addressed in day two of the workshop, where we’re talking about how to plan your stimulation, how to minimize energy delivery to critical structures, how to verify your exposure and monitor neural engagement.
Here I’m really calling on all of us to start reporting it so we have as much data as possible. But there are effects that are really very well reported and have been known for decades. Actually, those are the peripheral unintended effects of transcranial ultrasound. Especially pulsed ultrasound, it can be audible, you can hear it as an audible confound, that is the secondary, an unintended effect.
And at certain pulse durations and intensities you can also feel it. It might start very mild, as a tingling, then an itch, many people would know this is very similar to transcranial electric stimulation, especially TDCS, but at high exposure the itch might be to some sensations of heat, of warmth, of prickling, maybe even burning sensations. Now, in most of the human studies, we use far lower exposures. But I did not want us to ignore these peripheral secondary effects.
In fact, in the second session, we already heard a little bit about this, one of the most robust effects of ultrasound in humans is reduction of the motor cortical spinal excitability, but this is with the pulse protocol, pulsing 1000 times per second, you hear ding, the tone. So we wondered how much of this effect is driven by the primary neuromodulation, or a secondary auditory confound.
The first thing we did was actually put the transducer on another site of the head, another hemisphere, and that led to the same inhibition. In fact even playing the sound led to a preparatory cue. This is learned inhibition, you hear a tone, and you know the probe is going to come up, and this leads to an inhibition. And now you can see in this largest study that in fact we did not find any evidence of a clear primary effect, it was overshadowed by the secondary peripheral effect.
So we should all work very hard to minimize the secondary effects to be able to really start investigating and modulating what we’re after. That will have impacts for pulse shaping, or study design. But it’s not only to have the best study design and the clearest answers, it’s also because of safety.
Kim alluded to this earlier. In some conditions the tone can be painfully loud, creating some concerns for auditory safety. And the same I’ve highlighted here, high exposure on the scalp might lead to scalp pain, especially when the transducer is not properly coupled, and there can be a lot of reflections, and maybe even leading to skin irritations.
This isn’t systematically studied yet, nor is it fully included in many of the reports. Perhaps for some cases it is already common knowledge if you’re an ultrasound transducer, everybody knows that some protocols are audible and others are not. But if you’re new in the field it might be very hard to share this common knowledge and start to quantify it.
So there are some considerations for peripheral adverse events. Of course, there could be thermal heating, we heard about that in the previous session, through thermal damage. But I’m also talking about physiological safety, maybe skin redness because of peripheral vasodilation or nerve stimulation. With ultrasound it might also be because you’re putting pressure on the scalp with a transducer, there’s no cooling from the air, could there be risk of skin irritation that would also be present for SHAM.
So now we’ve moved from reported to common knowledge, and that brings me to a third level which is even more in the periphery. There are reports, maybe even anecdotes of adverse events. There could be dizziness from subcortical stimulation, or nausea, or even heart palpitations. We have never seen them reported in any studies. And you can also see that for the 704.
But that also means we will now have many more stimulation participants joining, and we shouldn’t ignore it, because actually it should be expected. If we have true neuromodulation in critical deep brain structures, and we already have all the evidence from animal studies where effects are very pronounced and strong, then we also can start expecting secondary effects, either on- or off-target, that will have clear relevance for subcortical structures. Think about motor control, your balance for the thalamus, or your breathing in the brain stem. Some people might get very worried when anything seems possible.
I also like to highlight that transcranial ultrasound has already been applied thousands of time. This is a very differently shaped pulse shaping, but for high intensity focused ultrasound, for thermal ablations, about 10,000 patients have been treated, where intensity is much higher from continuous wave, also for a longer duration.
And these are clinical interventions, clinical studies that are carefully monitored, and there is no seizure reported, even at these very high exposures, and in sensitive population. And in fact there are no clear, long-lasting physiological changes when you’re unable to reach the ablative temperatures. So that gives us some confidence on what we can expect for transcranial ultrasound at low intensities, as non-ablative intensities.
I also want to highlight our list of adverse events clearly will be growing the more participants we include. Some of them we can predict now. Namely we’ve already seen them in other non-invasive brain stimulation techniques. It can be quite scary to participate in these experiments, with all the apparatus and the machines around. If you haven’t had enough to eat, or it’s a hot room, there could be a risk of fainting.
But as it is for transcranial direct current or transcranial magnetic stimulation, this isn’t an adverse reaction. It happens, temporally it’s associated with these studies, but not directly induced by the stimulation, and I’m sure that in the future this will also happen for transcranial ultrasound studies, but I have no evidence, neither empirical nor theoretical suggestions that fainting could be induced by ultrasound.
And similarly, no doubt we’ll be seeing in clinical studies symptom deterioration as we also see for other NIBS techniques, like a worsening of tremor or a worsening of depression symptoms. Now, for those who want to take a screenshot, here is the full slide, discussing some things that are already reported, some things that we might expect in the future as we take into account.
But I would like to take one step back and not only have a retrospective of what has happened to date, but also a prospective, what effects are possible or likely. Can we make a principled assessment of risk and safety prospectively, and I would argue that therefore we would need to consider the physiological range of the intervention.
So I am going to set out a look at electromagnetic stimulation, and I am going to sort them on a scale of the intervention from completely reversible changes to irreversible changes in things they were doing. If you have an irreversible change you have a structural change, and your structural effects. And the most reversible are where you have a modulation of the natural variations.
And if we were mapping current electromagnetic stimulation techniques, we would have high intensity ablation at the very top. And we can already see that there could be a mixture, for example for electro-convulsive therapy, or much more reversible effects in transcranial magnetic and transcranial electric stimulation. And these are separated because transcranial electric stimulation at low voltages is able to clearly evoke synchronized activity, while TMS is.
While we are looking at what the effect depends, for ablation it’s often only on the biophysics, but all the way for reversible changes we need to consider the neurophysiology, the effects that are dependent on neural activity, and a whole wide range of conditions and parameters.
And that means that the major consideration of reversible neuromodulation is often not effective at all, while for example for electroconvulsive therapy we carefully considered a therapeutic window, we want to have robust clinical effects, but don’t want to have the dose so high that we might have irreversible or unacceptable irreversible damages, that’s the therapeutic window here.
Mapping this approach to transcranial ultrasound, I would actually argue that we have a similar set, ablation for structural changes, blood-brain barrier disruption for extra physiological changes, and neuromodulation at the moment really seems to be spot on here in a modulation of activity within the physiological range. So efficacy is one of our main concerns, and underdosing is one of the challenges.
So, what can we expect in the future? I would argue that we should be asking questions like are we modulating spontaneous activity or are we evoking synchronized activity. And if that happens, is that within or beyond the physiological range. For example, if we see a reduction of GABA, is that a modulation that would happen in a natural variation as well, or is the reduction in GABA so strong that it is outside of the physiological range? That is the moment we can start expecting some adverse events.
We can learn from other brain stimulation techniques. Deep brain stimulation, where we have a low risk of acute seizure, but there are very clear acute side effects for speech, walking, muscle weakness, prickling sensation, or even risk-taking. And because this is such a synchronized evoked activity with chronic long-term use, it is known in the longer term to lead to neuronal damage for example through over-excitation.
Moving to transcranial magnetic and electric stimulation, these are quite distinct in the sense that TMS really elicits brief, synchronized activity. TMS drives the neurophysiology. That creates a risk of extra-physiological activity, such as seizure.
Now transcranial electric stimulation, the low voltage and moderate current here, that modulates the neurophysiology, it changes the probability of spontaneous action potentials, and actually there is no known or negligible risk of extra-physiological activity. That can be contrasted against the high voltage and low current TES that can drive action potential.
So a critical difference is do I drive action potentials directly by the stimulation. This has been studied in animal models for ultrasound. It’s clear that ultrasound has a delayed onset latency, for example at least 40 milliseconds. The resulting modulus of activity is not highly synchronized, it is very dependent on the spontaneous activity, and in low to moderate intensities there may be some modulatory effect. Who knows at higher intensities, if we move to domains that aren’t often explored and may not even be biophysically safe to use, could synchronized activity be possible.
We had a fantastic talk by Shy Shoham where we learned that both inhibitory and excitatory neurons respond to ultrasound, so that matches a lot of what we heard earlier from Keith. Maybe inhibitory neurons have a lower threshold. They maybe dominate at the low intensities, while at the higher intensities maybe the excitatory response could dominate. And Shy, I screenshotted your presentation to include these wonderful figures. Thank you very much.
But I wanted to highlight that there isn’t a one-on-one mapping from animal studies to humans. Most of these studies often rely on calcium imaging, that is not the same as spiking activity. Similarly, maybe most critically, there is no TUS-evoked muscle twitch in humans. We can learn a lot from what we already see in humans, right? Just to repeat, I actually do not know of any clear study that has specific TUS evoked potentials with appropriate control conditions. I do not know of any evidence of neurostimulation addressed when the neuroactivity is at risked, no evoking of potential. There is clearly no evidence of highly synchronized activity as it would seem for TMS, and so far no hard evidence of any extra-physiological reaction. Instead, we have strong evidence for modulation of task-evoked potentials, and task-active regions.
So, at the current protocols, the current exposure levels used, TUS modulates spontaneous activity, but it doesn’t evoke synchronized or brief synchronized activity. Maybe currently it mostly matched to what we can expect from transcranial electric stimulation like tDCS or tACS or tRNS. But will it always stay within that range? If it does, that can give us a good confidence. Stimulation at one intensity might lead to a facilitatory change depending on the stimulation. Dose, we might see long-term potential as indicated here by the arrow.
But as often happens if you’re stimulating within the physiological range, repeated stimulation means that actually the dose-response plasticity curve is changed, the same stimulation a second time around or repeatedly might lead to an inhibitory response. That’s a homeostatic effect, that means there are no runaway effects.
We must acknowledge that most animal studies currently focus on excitatory and inhibitory activity. So immediate acute modulation of this activity. While many human studies are using offline designs to study short-term plasticity. But still looking at excitation and inhibition. And in the future, I would not be surprised if we are looking more into gating of plasticity, or introducing neural noise without a specific E/I change. It is tricky in fact in the healthy range to modulate this, because it’s such an optimized system.
We are often already at a peak performance, in the Goldilocks zone, where any change in excitability or any modulation might bring you away from the optimum and lead to a perturbation of function or behavior. If you were using in a disordered state or diseased state, there might be much stronger modulatory responses, even to the same stimulation parameters.
Now, there are many open questions. And for example, about the strains of what would happen if we are exploring more protocols, moving beyond what we currently do. How do we define exposure and dose. How do we define the therapeutic window. Some of them will be discussed in this panel, and others will be addressed in tomorrow’s sessions. For now, I would like to thank you for your attention and my wonderful lab for providing all of their thoughts, ideas, and work. Thank you very much.
Clinical Safety in Mental Health Therapy
FIDEL VILA-RODRIGUEZ: Thanks so much for the opportunity to present today. I will take advantage to introduce myself a little bit. I come to the workshop from Vancouver in Canda from the University of British Columbia, and the laboratory here at UBC is specialized and has a particular interest in any neuromodulatory interventions that can help alleviate the suffering of people with psychiatric disorders. So we’re going to be embarking on low-intensity focused ultrasound in the very near future, and it is a really excellent opportunity to be attending the workshop today.
Again, I was talking about my disclosures, and I was saying that I don’t have any conflicts of interest, specifically with focused ultrasound. The lab has received funding from different funding agencies. And we do have received in-kind support from MagVenture with magnetic seizure therapy device for an investigator-initiated project.
I like to start with a bit of a roadmap. And I will be taking a lot from the speakers that have already presented, and I’m hoping to provide some of an overview and a roadmap of organizing the content that has been discussed so far, and how it leads to clinical applications. This conceptualization of non-invasive neuromodulation technique is applicable to any type of neuromodulation.
And we always start with biophysics considerations, and whether we can design devices that can transfer energy, and along those lines in this stage we consider a lot of aspects around biophysics and modeling. Once we have a device that is able to transfer energy, then the question that follows is about the biology.
And here is where earlier today, we’ve seen what are the effects of this transfer of sonic energy onto tissue and cells as a purview of that basic science. Once that’s established then we move on to kind of a different level of questions. In one case Leonard has discussed applications of focused ultrasound for the study of how the brain works in healthy volunteers.
If we apply focused ultrasound for the purpose of treating a medical condition, then the questions that we deal with are whether those biological effects actually improve outcomes, or as well we’re very interested to know whether the intervention with the focused ultrasound is associated with any adverse events. This is the purview of clinical trials. And focused ultrasound, low intensity focused ultrasound is in a very exciting stage. We have heard a lot about the excellent and very thorough science on biophysics, on basic science.
And right now we are at that stage that we can confidently start to think about applications, therapeutic applications. There are not a lot of clinical trials actually published, let alone in the space of mental health. But there are an increasing number of clinical trials registered, and I just looked at one of the clinical trial registries, and there are more than 30 at the moment, so this is something that is coming up, it’s something that this workshop comes at a very timely point in time, in the development of low intensity focused ultrasound.
Similar to what Leonard just did, I would like to pause for a second and go over a few standard definitions that are adopted in the context of clinical trials that will help us anchor the discussion with the panel and also hopefully provide some framework for us collectively in the field to reflect about how we go about ensuring safety in clinical trials.
The first one that is very important is adverse event. In clinical trials an adverse event is an untoward medical occurrence or effect. And I put in italics untoward, because those of you who shared with me that English is your second language may or may not be familiar that untoward is a beautiful word with very nuanced meaning. It has the connotation of something negative, something inappropriate, inconvenient.
So an adverse event is something negative, something inconvenient. It could be a medical occurrence or effect that is either subjective, that is reported by the person, I am feeling pain, I am having headaches, or it could be something that can be objectivated, for example some bloodwork analytical parameter that goes off.
And so those are some examples of adverse events. It’s important to note also the nuance of the word event. Event is like anything. It’s very open. And it's very also important to know that in the context of clinical trials, adverse events are monitored during the entire duration of the study. That’s a particular area that sometimes we lose site of it.
But when a person enters a clinical trial, and that is the time when that happens is very concrete, is when they consent to enter a trial, they enter kind of a special dimension in terms of the team, the research team is monitoring, is collecting data about what is happening. So anything that happens is recorded.
And what I am getting at is perhaps best illustrated by an example from one of our clinical trials, where someone consented to enter a clinical trial to use transcranial magnetic stimulation for depression. And after the day that we did the consent and did baseline measurements, and a week later when we started the actual intervention, in that time they went over the weekend for a hike, and they kind of twisted their ankle.
As far as we are concerned, we record that as an adverse event. It is something that has happened during the clinical trial in that person. Later on we will discuss about the causal attribution to the intervention. Obviously they did not have any TMS, so it is extremely unlikely, it is not related to the TMS. But I think it is really important to keep that context in mind.
We usually grade the severity of adverse events. And usually we use terms like mild, moderate, and severe. It is really important, and again the language here is very specific in the context of clinical trials. A severe adverse event is something very different from a serious adverse event in the context of clinical trials.
Serious adverse events are adverse events that are life threatening or result in death, that require hospitalization or prolong an existing inpatient admission, or adverse events that lead to medical or surgical intervention to prevent life threatening illness. So it has a very particular meaning, and is dealt in a very specific way in terms of reporting and how to report them.
The other connotation in terms of adverse events is whether they are expected or unexpected. And that dimension speaks to the qualifier of whether we have what are the prior expectations about the likelihood of something to happen. Again, it’s expected most likely that it could happen, that some people might have a headache after TMS. We want to define and we need to really understand better in the context of low intensity focused ultrasound what are those prior expectations.
The last term sometimes gets conflated with adverse events, and it is a side effect. A side effect is a concept more in the purview of clinical practice. And side effects are potential negative effects of interventions that are likely to be causally linked to that intervention, to that treatment.
So when we go about doing clinical trials, the information that we derive from them eventually inform on the label of interventions what are those things that potentially have a chance to happen that are more likely to be causally linked to that intervention. That is different from an adverse event. If someone would like to know more about all these definitions and the latest update from the CONSORT group, I put the citation down here on the slide.
Now that we have definitions, I would like us to consider what are the different levels of side effects. And for this I put together this slide that talks about three emerging levels of safety. And in trying to put together a lot of things that have been discussed today, and people, and there have been talks about biophysics aspects of safety, we’ve had talks and speakers referring to nonhuman models of safety and what they look for.
And right before this talk Leonard talked about studies, adverse events in the context of human populations and healthy volunteer populations. I would like to conceptualize the emerging levels of safety and what do we do as a field to safeguard the safety of human participants. So this is like a raft that protects us from untoward things, kind of the sharks around in the water.
And the first level, which is extremely necessary, but still not sufficient, is all the things that we’ve learned today about the biophysics of low intensity focused ultrasound. We’ve heard about considerations about thermal effects. We’ve heard about the importance of learning about the acoustic focus, mechanical effects.
And we’ve learned about what are those parameters that establish a bit of a safety boundary. That’s a way that I find it helpful to conceptualize those. Those are kind of boundaries around which we establish okay, this increases the safety of what we are doing. I want to emphasize that slight does not allow me to convey the sense that these boundaries are not fixed in time, and they evolve as we accrue more new knowledge. So what are the safety parameters today as we accrue more information may or may not vary.
All right, we covered the first level. We do our due diligence, we get MRIs, we get CTs, we model our target, we have that information of safety parameters. The next level are the studies that a lot of folks in this group have been talking about. These are nonhuman models. We’ve heard about sheep, rodents, nonhuman primates, and in those models what we can look at is within the safety parameters or outside of them what are the effects on tissue, on cells. So that is pathology. And are there any histological findings. We’ve learned today, earlier today, about cavitation, about different potential types of lesions at different levels of intensity. That’s the second level.
We have increased our quantum of information about our certainty about this safety net that protects us from causing harm. And then we go to the level of trials and studies in human populations, in healthy populations. I want to convey that the information gain in both populations can reciprocally inform safety.
So as Leonard rightly pointed out on his talk, all the experience that has been accrued by researchers who have conducted low intensity focused ultrasound in healthy volunteers do provide invaluable information that added to the biophysics and the nonhuman safety models make us more confident of making that step that leads to start doing clinical trials in humans in clinical populations.
And again, in the context of clinical trials, it is extremely important to always monitor, report, and record adverse events, whether they are subjective or objective. And this is an iterative process. So as we do more clinical trials, as the field advances with more information in nonhuman models and biophysics, we update the information here, and treatments become increasingly safer and establish the parameter space.
Tomorrow we are going to hear about regulatory aspects. And I am not going to go into the details about regulatory aspects, but I want to introduce how the three levels of safety that I just discussed play out in the sequence of discovery, specifically in the field of low intensity focused ultrasound. The sequence goes such that the field has done now and accrued a great deal of information in preclinical biophysics, and so we’ve established some parameters on safety.
And also the second level of safety that I just showed, these studies of animal models and safety, these are critical input, along with information on healthy volunteers. And then we enter the formal phase of clinical trials, investigation in humans for the purpose of therapeutic effects. Each of these phases it is important to note that they are indication specific.
So what we establish in this pathway for major depressive disorder needs to also be established for anxiety disorders, for schizophrenia, for any indication. It's specific. That’s a very important point. Studies in phase one are initial studies, are early on, where the emphasis is placed a lot on safety.
We really want to get a really close look at what happens with people who receive low intensity focused ultrasound the first times that we do this. They are usually small sample size studies, and usually are very helpful to interrogate questions about dose finding, which we’ve heard today that the concept of dose in LIFU as well as in other non-invasive brain stimulation therapies, it’s very nuanced and has a lot of parameters that factor in to consider what is the dose delivered.
Questions about efficacy are still there. And so we also monitor whether LIFU is associated with good effects. But the onus early on is more on safety, which it never disappears, we always look at safety throughout all the lifespan of an intervention. But the emphasis starts to shift slowly. In phase two, if we’ve demonstrated that there is reasonable safety, we can progress to start recruiting more people into clinical trials, our larger sample.
They investigate questions about is this feasible to do. What is the dose response? We start to get signal about efficacy that allows us to plan the next stage, which are pivotal trials, or studies in phase three, which are large trials, the goal of which is to establish that the treatment works. We never lose sight of safety. Safety is necessary for approval but is not sufficient. What we want to show, and what leads to this boundary here and to a treatment to move into implementation is that we know about its safety parameters, that it is acceptable, and that it works for the indication.
At that point, interventions, moving to phase four, where the questions that we ask continue to monitor safety. It’s very important that never, ever stops. And there were questions earlier about the audience, do we know anything about the safety of LIFU in youth, do we know anything about the safety of LIFU in pregnant people, other special populations. What are the considerations in terms of health economics and implementation.
Those are all questions that are geared in phase four, and that kind of look primarily at effectiveness, but also pay attention to potential additional adverse events that might emerge as we treat more and more and more people that did not appear early on in the trajectory of clinical trials. Again, tomorrow we will hear a lot more about regulatory context, and we will come back to the different phases in clinical trials.
In closing, I wanted to also give some practical considerations. These are thoughts to hopefully stimulate the panel discussion to get the audience to get into the mindset of what we might want to do in the context of LIFU, to make a robust entry into the realm of clinical trials. The first and most important aspect is that we systematically monitor, record, and report adverse events in clinical trials, from the day that people enter into the trial to the day that they exit the trial. That never stops. We are watching, we are recording. We have the duty to report them.
Oftentimes a combination of open-ended questions about potential adverse events with standardized questionnaires is optimal. And so it would be really neat to kind of establish and harmonize a little bit what those potential standardized questionnaires are in the context of LIFU.
Then once we start monitoring and recording adverse events, which is the first step, the second step is to make those inferences and to try to tease out and ascertain what is the causal link. It is always a degree and a manner of how much certainty between adverse events and the actual effect of LIFU. In other words, which of the adverse events that we record in trials are going to make it onto the label and noted as potential side effects of LIFU in clinical applications.
I think it’s important to keep in mind that regardless of the prior likelihood that we believe different adverse events might have, it’s important to always ask and pay attention to very common adverse events that are reported in any other intervention, things like lightheadedness, headaches, pain or discomfort at the site of stimulation.
Another one that I didn’t list here but comes to mind now is fatigue. So those are very important to record, and to ask about in a systematic way. But it is also important to be alert for uncommon adverse events, for things that we might not expect but do happen. We need to be very fine-tuned.
Other speakers have talked about the aspect of hearing safety. Tinnitus is a condition where there is auditory perception of usually high-pitched noise, not related to auditory stimuli, and it is very common. So I think the field would be important to keep tabs on whether there are any potential adverse events in the context of LIFU.
In the context of psychiatric indications and condition specific, it is really important to consider in terms of adverse events the adverse event of worsening of the condition, whether it’s worsening of anxiety, mood, or very important, suicide ideation. We know that we have to pay a lot of attention on these domains.
Something that has been discussed earlier today is that it is important to keep in mind that there might be indication specific adverse events. SO there might be things that are really important for one indication, but are less so in another. So we need to keep our thinking flexible along those lines. And it’s also very important to keep in mind that there might be individual characteristics that kind of moderate or change the likelihood of adverse events. Things like age, ethnicity, or other parameters.
And my last slide is perhaps to steer a bit the discussion, language is really important. And something that we learn a lot in clinical trials is the aspect of nocebo, which is kind of the reverse aspect of placebo. Nocebo are nonspecific negative effects that can happen due to participating in a clinical trial.
And as a clinician, as a trialist, the name low intensity focused ultrasound has particular appeal because it feels to me that it’s very descriptive, it really contrasts with another use of focused ultrasound which is the ablation application, and it might, as a reflection that I throw out there, perhaps using this type of language potentially minimizes the nocebo effect.
Thank you very much for your attention. I look forward to the discussion now. And I want to acknowledge the support of funding agencies and colleagues that helped. Thank you so much.
ELIZABETH ANKUDOWICH: Thank you Fidel, for some really interesting clinical trial considerations. So we have a minute or two for questions. There was one that came in during Fidel’s talk. I think it relates to some of Lennart’s presentation.
So the question kind of talks about motor effects, including those presented earlier by Dr. Aubry and kind of countering that with some of the findings you’ve seen in humans where there’s no evoked motor response. And so the comment is, is this possible if one were to properly compensate for potential additional barriers?
LENNART VERHAGEN: It’s making an excellent point, going back to something that Jean François shared with us in the earlier panel discussion. So let me recap. We’ve seen from Shy Shoham’s talk how strong effects in animal models can be, how convincing and how robust. In human studies are much lower doses, so the effect might not be as robust.
But Jean-Francois was highlighting some recent results where there was a very strong and substantial reduction of tremor when there was very specific targeting of low intensity ultrasound in essential tremor patients before they move for ablative high intensity focused ultrasound. Maybe this was in response to arguments I’ve made that there is no clear evidence of highly synchronized evoked activity by ultrasound in humans.
Actually, I will completely agree, these two things are entirely in line. The modulation, the reduction of tremor, is a very strong example of modulating spontaneous activity. And that is where we have a lot of evidence that ultrasound does it, and this is particularly striking evidence in humans.
Secondly, I’ve been arguing that such perturbations might even be stronger or more expected in diseased state, when actually the circuits are disordered, not more optimized, then we might even see stronger effects here of ultrasound, that’s what we’ve seen before.
So I’m very excited about the work that Jean-Francois was talking about, and it also shows how carefully you need to target and focus your ultrasound and titrate your dose. Then the effects of modulating spontaneous activity are very striking and robust. Thanks.
ELIZABETH ANKUDOWICH: Is there a parameter range that might be above the diagnostic guidelines that could evoke synchronized activity without inducing tissue damage? Because that would be a very valuable tool, similar to what we have with TMS, but at depth.
LENNART VERHAGEN: To kick off, currently we are often already crossing diagnostic guidelines, especially for the intensity, the spatial peak and temporal average intensity. TUS protocols have longer pulses than diagnostic ultrasound. So quite often we’re already crossing those diagnostic guidelines.
The question here highlights an important part, can we envision future protocols that not only modulate activity, but really can evoke this as we might already have evidence from in animal models. I am very hopeful.
We see from the animal literature that specific protocols can be extremely effective. Many speakers have already highlighted how complex the problem is, how many parameters there are. But also within those parameters and different targets, maybe there are great matches that are particularly effective.
And this would require us to really do parameter sweeping, to carefully and systematically map this out. We shouldn’t be too afraid to see evoked activity. So having evoked activity with ultrasound, these highly effective protocols, would be wonderful. It would be a fantastic tool to have deep brain TMS without superficial stimulations. I hope we can move in that direction, but we’re not currently there.
SHY SHOHAM: Lizzy, in response to your question I also want to mention that all of the work in ultrasonic neuromodulation in humans we not with phased array, maybe with the exception of Jean-Francois’ result. And in humans we have this highly aberrating skull. So we should highlight actually the work of Yan Kubernick(ph.) himself, who is kind of spearheading the application of phased arrays to humans, and that could potentially be an important step and a way to more evoked activity in humans.
ELIZABETH ANKUDOWICH: Thanks for that response. We did have one more question related to minimizing placebo and nocebo effects when kind of designing these clinical trial protocols. I’m wondering whether that might be a good starting point for some of the other clinical trial considerations that you might want to discuss with the panelists, including ways to enhance efficacy and minimize adverse events. Leonard, do you want to pick up from there?
LENNART VERHAGEN: Maybe we could first listen to Fidel, how you feel about this question. And I believe there is one more question popping up in the chat.
FIDEL VILA-RODRIGUEZ: Of course. And with regards to placebo effect, and again a reminder these are nonspecific positive effects, the main one that has been brought up today and that we need to keep in mind is the auditory aspect. Even though the ultrasound is out of the range of human hearing, still you can know that something is going on.
So the design of a good sham that mimics that sound but does not actually deliver the energy is one of the key aspects here. And there are already people using sophisticated ways of doing sham, such that the sound is active but the ultrasound does not focus and deliver the energy at the point. So that is for the placebo, nonspecific good effects.
To minimize nocebo it is really important to have clear expectations about what the treatment is about, and resolve areas of uncertainty and confusion. And by that I mean, I will bring an example with TMS. Still to this date one question that I get many times from people who are considering TMS is, is TMS ECT? So we use electricity, and still sometimes people are wondering, are the side effects of TMS similar to ECT.
And clarifying those aspects ahead of time, and knowing about what are people’s questions does help a great deal, because imagine if we had left someone with that question when entering a trial, are they expecting that they are going to have similar side effects as with ECT? That could lead to that nocebo effect. I’ll stop here, and perhaps get to the panel.
ELIZABETH ANKUDOWICH: I think it is time to transition. We’ve run out of time for questions. So if you want to take it from here Leonard, that would be fantastic. Thank you. Thank you to our speakers.
LENNART VERHAGEN: Thank you all. Let’s move to the panel. I would like to invite all panelists to wake up their screens. We have also had a few questions posed in the chat, I would love to bring those wonderful discussions in the chat also here to the public panel discussion. So it would be great if we could bring that all to the front. But I’m actually going to start with a question that was posed earlier.
The question here is might glial play a more prominent role, for example owing to mechanosensitive receptors in TUS compared to transcranial electric stimulation, and if so could this confound or limit how much ultrasound can learn from other stimulation modalities. That’s an important question. I want to get started, maybe Til, you have used a lot of noninvasive brain stimulation studies. How much can we learn from other studies on the primary or maybe even secondary effects of ultrasound?
TIL BERGMANN: I’m Til Bergmann from the University of Mainz in Germany. I’ve been doing TMS and TUS for quite a while, before we also got into the ultrasound business, and I think there is a lot we can learn from the other techniques. I can’t tell you too much about how much the (inaudible) effect might be related to glial cells.
It’s actually known that some of the TMS effects, especially the artemis effects, repetitive stimulation have effects on glial cells. There is some pretty interesting evidence of their contribution on the long-lasting effects. So it’s definitely something that we should consider, especially for the after-effects, to look into.
A point I would like to make with regard to the secondary effects, that’s actually the same for a focused ultrasound and all the other non-invasive, even invasive techniques in principle. In previous talks, which adverse effects can occur, and effects can be due to (inaudible). But even if we just consider those who believe they are caused by stimulation of the brain tissue, there are still at least four different types of secondary effects.
The first would be that we stimulate our target neuron population in the target circuitry that you want to stimulate, but maybe with an unintended strength or an unintended direction of effect, or unintended type of effect, so that we get a response that is undesired. An excitation instead of an inhibition, with synchronization I don’t think this is really happening with focused ultrasound, but in principle.
The second would be that we’re stimulating still our target region, but we’re affecting neurons that are not part of our target population. So neurons that have different connectivity fingerprints, that have different characteristics that actually serve a different function, because they’re part of a different connectome.
Third, off-target stimulation basically of neighboring regions, because as we’ve heard typically with (inaudible) elongated focus, and there’s no way this by chance fits exactly our target structure shape. Plus we don’t know exactly what the threshold is for evoking neuromodulatory effects.
So even if we give this full width half maximum for our focus, it doesn’t mean anything with respect to what is the effective size of the simulation. So we may stimulate functional circuits neighboring that have a completely different effect on a behavior level.
And then finally there is the potential stimulation of non-human elements, glial cells, maybe direct effects on the vascular system, that is really not much known about, but that should be studied well in animal models first, in the future.
LENNART VERHAGEN: Dr. He, you’ve done a lot of studies on ultrasound, both in animal and in humans, learning about, seeing so many primary and secondary effects, what do you consider the risks, were there any that you maybe haven’t heard yet in these talks, what neurophysiological risks should be considered?
BIN HE: I think there are lots of points that have been covered. Certainly the most important, you look at the mechanical damage or thermal damage, and you measure the temperature, look at the structural damage. I wish to just make one point. I would suggest that also lots of discussion in this chat among the panelists, Holly brought it up. I would add the neurophysiological monitoring and assessment.
And I want to add more about what we communicate. We did observe that actually in small animals, in rat, we were able to give focused ultrasound stimulation, we recorded EEG, we were able to localize the source.
Very recently, in unpublished work, there is an experiment in monkey we just completed. We did a similar experimentation by putting focused ultrasound on front eye field, and the way the monkey has eye fixation, and then with ultrasound (inaudible) and we observe that in the waveform we are able to see something like evoked potential if you would average these phase-locked activity.
We were also able to localize the source, which we haven’t completed the work, so we don’t know exactly how many or what is the accuracy, but it is clear that it is possible, Lennart as you pointed out, like PMS or others, the focused ultrasound actually is inducing electrophysiological effect that can be quantitatively assessed. Of course, more work is needed. So my point is we should add electrophysiology as one of the multiple tools assessing the efficacy and the risk.
LENNART VERHAGEN: That brings up an important point. And we’ve also had it in the questions the audience has asked. There has been no observed direct reaction of seizure with ultrasound. But Mark, you’re not part of this panel, you were part of the previous panel. Would you mind popping back up? Did I hear you correctly, in one of your studies there was an adverse event of a seizure?
MARK SCHAFER: The study was in fact a study of exposure in epilepsy patients. We had eight patients who were going in for a resection, so they were long term severe seizure patients. We had one patient, I was recalling this, it ended up not being the study because the supervising physician determined that the seizure which happened several hours after the treatment had ended was typical of their pattern of seizures, it was not considered out of the ordinary. We did pre- and post-MRIs on these patients, and also a neuropsychological assessment, those all came up clean, as well as the H&E later.
I just wanted to clarify. I was remembering anecdotally yes, I’m looking over the entire record one did, but it was determined that was well within their normal rate of seizures. These were all people who were refracted to treatment otherwise, and that’s why they were going in for resection surgery. So it wasn’t as that we caused it, it was just part of their clinical workup.
LENNART VERHAGEN: Still an adverse event during the trial, but maybe not trial related.
HOLLY LISANBY: Thank you all for your really informative discussions. Thank you Mark for clarifying that what you had mentioned previously was not likely to be a seizure induced by the focused ultrasound. What I heard Lennart from your talk is in the human work so far that there has been no clear evidence of focused ultrasound evoking potentials in humans, and no clear evidence of neurostimulation at rest.
Often we use the words modulation and stimulation as if they were the same thing, but I think a distinction is being made here, that you’re saying in the human work so far focused ultrasound has not stimulated an evoked potential, but modulates ongoing activity.
But we also heard some really important animal research, and Bin you just mentioned that in your work, and others are really important, that in some parameters in some species there does seem to be evidence of focused ultrasound being able to evoke synchronized activity, not just modulating ongoing activity. And I think this could be useful as a research tool. We really would like to be able to focus stimulation at depth. If that could be accomplished in a therapeutic window where you’re not causing tissue damage, you are causing stimulation.
And if such a tool were available, since we currently have that for cortical stimulation with TMS, if we could do that at depth, then we certainly would want to know the seizure guidelines, because if you are evoking stimulation then you could induce a seizure, and that would be the space where we would want to know what are the parameter sets that we need to stay below to avoid that risk. Thank you for letting me parachute into your panel. Back to you, Lennart.
LENNART VERHAGEN: I am very thankful that you did because these are fantastic and very important points. On the panel we have quite a few experts who might want to weigh in. But I’m also hoping to learn from Ellen, Holly was just describing maybe at the current stimulation settings and parameters there’s no evidence of seizure risk. Is there a possibility that we might induce a seizure? what would be needed for this, and how do we know where that transition is, and what should we be doing for monitoring?
ELLEN BUBRICK: Thank you. I’m Ellen Bubrick, I’m an epilepsy neurologist, clinician scientist. And these are all really important questions. So I think it is possible that at certain parameters it could cause seizure. It is possible that some drugs at certain doses that don’t normally cause seizure cause seizure. It’s possible that a lot of things that don’t normally cause seizure could cause seizure. So I think the question is why would we need to look at it. I’m less worried about it causing seizure when we have thousands of normal human controls where there was no seizure reported.
But I do worry, as Holly was saying, she’s absolutely right, we do need to do that parameter sweep, we need to know how high is too high, and what other problems could we start getting into. Just like you do the toxic dose limiting studies for drugs. Of course we will need to know that. We may not need to get that high actually.
We’re already seeing some significant evidence of neuromodulation at fairly low intensities. So I don’t know that we want to play with fire, go too high. We just want to know is it safe, to get the kind of neuromodulation we need to modify disease. So we may never get to a point where we see seizure. But we may see lots of stuff at higher doses that we don’t want to see, and that’s how we back down.
LENNART VERHAGEN: There might be some clinicians and researchers who actually would see a tool that can induce focal seizures deep in the brain as a very valuable tool. But moving to future clinical applications, in many cases what we’ve seen in the past is that clinical treatment might require repeated stimulation and repeated sessions. What do we actually know about the safety of repeated sonic stimulation? Miriam, would you mind shedding some light on this?
MIRIAM KLEIN-FLUGGE: Sure. I’m Miriam Klein-Flugge, I’m a neuroscientist at Oxford. I think it was maybe kind of hinted at, but currently there’s actually very little published data, at least in primates, that administers repeated or chronic TUS or ultrasound stimulation.
But there are two studies in macaque that Elsa did earlier in the first session. And so they have more than 100 sessions in these macaque monkeys, one study was with even daily sessions over six months, and they do not flag any safety concerns in those animals. That’s obviously based on behavior that we can measure from macaque monkeys and MR based measures as well. And that is also consistent in general with the rodent literature.
But of course, it maybe doesn’t necessarily tell us about how tolerable it would be to humans. But here I think we have some great parallels with the TMS world, and we know there are some treatments that have been FDA approved for years that have quite tolerable but intense repeat sonic stimulations in that case, like the recent protocol that came out of (name) lab.
So I think I would say further careful kind of characterization using MR and behavior and physiology, and maybe standardized reporting of any secondary effects would be really useful in humans with repeat sonications and multiple sessions in the same individual. But at the moment at least from the data we have it does not look like we should be worried, that there’s any reason to think that structural damage or secondary effects are more likely to occur with repeat compared to single session administration.
LENNART VERHAGEN: Thanks. Noah, you are running a clinical trial with transcranial ultrasound. Would you might shedding some light on what specific considerations or maybe contra-indications we should take into account?
NOAH PHILIP: Noah Philip, a psychiatrist at Brown University. I think, to echo some of the things that have been said previously and maybe build upon it, I think first, we are doing an NIMH-funded study of focused ultrasound to the amygdala. And oftentimes when we’re thinking about this new technology we’re targeting areas that have never been modulated non-invasively, when we’re talking about deep brain regions.
And so for example, in our work we rationally and intentionally chose not to use a bilateral administration. These were thoughtful about what those side effects might be if we went in that approach. And I think thinking about the effect of what it is that we might be doing are really worth thinking about, and really being very cautious about adjacent structures, and being thoughtful and sensitive about evaluating for what we would call off-target effects, but really are the beam going somewhere else.
The only other thing I would like to add is the field is very new, and we have heard lots of work to this effect today. But even terms like excitatory or inhibitory or stimulation, we’re still really new in this field, and those words have specific weight and directions of effect. And we’re still not really quite there yet with understanding the direction of the effect of these ultrasound applications of various parameters. And so as we go into studies, doing in humans, we just need to be thoughtful, being able to measure whether things are going up or going down.
LENNART VERHAGEN: Since we have such a fantastic panel here, we’ve learned a lot about everything that we should be considering, what is expected or maybe unexpected. I would love to learn from you. How are we going to tackle all of this? So what are our best practices? And I realize we only have a few minutes left here in the panel. Maybe I can ask all the panelists to give a brief best comment. Maybe Miriam if you could kick off best practices, how can we minimize side effects or reduce burden.
MIRIAM KLEIN-FLOGGE: I will try and keep it brief, but that is obviously a huge question, so there are a lot of things one could possibly say. I think sonication goes beyond just mechanical and thermal safety, but for example you’ll notice that the sinuses are very different in different people, and if you hit it you might actually induce more pain and sensation.
So I think careful planning also impacts that side of things, the secondary effects potentially and as we already mentioned off-target undesirable side effects caused by off-target sonication. I think it has some really simple precautions that are probably trivial to even be mentioned, but we always tell participants to get a good night’s sleep, to stay hydrated, to eat, to have no alcohol, no recreational drugs. It’s very straightforward, but you will notice that some secondary effects that they will report do go down based just on some very simple checks.
And then during the session I always like to think of it as making the hairdresser experience, because I think anything that helps the nerves really helps. It sounds again very trivial, but if the participant is comfortable and not nervous it does help. And I think (inaudible) is not possible because they’re not in a scanner. We’ve already heard possible auditory problems.
LENNART VERHAGEN: We should devote a whole session to this. Noah, are there any best practices from a clinical perspective?
NOAH PHILIP: Two things. And for those folks who are in the audience and are used to working in a brain stimulation space, here thinking about the school is much more important. And we might even need to start thinking about conditions that affect the thickness of the skull, the composition of the skull, or perhaps even connective tissue disorders and things in that space. And I think frequent sampling is going to be very important as well, since we don’t quite have a good feel for the timeliness of the evolution of both on- and off-target effects.
LENNART VERHAGEN: Professor Bin He was arguing for careful assessment monitoring, especially including electrophysiology. Til, you’ve done a lot of non-invasive brain stimulation techniques. Are there best practices for assessment that you could share with us?
TIL BERGMANN: It depends massively on the effects you are expecting. You cannot monitor everything at the same time, so you have to choose your core modalities that best fit your effects of interest in a more sensitive way. EEG is definitely an important one, but you also have to remember very small amounts of tissue that we’re stimulating that might not even create a sufficiently large sound potential at the scalp to pick up with EEG recordings. FY(?) is another one, we’re going to hear more about it I assume tomorrow, which may also pick up more of the induced, less phase locked activation.
But I think it’s crucial that we think appropriately about where are we going to stimulate, where are we going to have off-target effects likely, and then based on our neuro-anatomical knowledge and acoustic simulations have an informed idea about which types of effects to expect when we’re stimulating close to the brainstem, and there will likely be an overlap, we may want to monitor some vital functions, some heart rate, some breathing, et cetera.
And estimating close to some limbic or basal ganglia nuclei, we might want to monitor some other functions like mood or a motor function. We tend to do by questionnaires. The last important point. And I think we do need in the community consensus and a standardized questionnaire, which will also hope to gather information for future meta-analysis and risk assessments. So this is something I think that we should work on together.
LENNART VERHAGEN: Let’s start tomorrow. Ellen, is there anything to add, for example on the assessment, from a clinical perspective, from the studies that you’re running?
ELLEN BUBRICK: The EEG part is so interesting, it’s so important, and yet so difficult. Til and I talked about this at Fund23 Conference, that when you have the transducer right sort of where you want to be recording, it’s very hard when your electrodes are sort of off-site then from where you’re sonicating. So it is hard to do EEG at the same time. So we have to sort of look at other ways to do it.
So EEG, I think it will ultimately when we can figure out a good way to monitor can be helpful. But it’s also just having serial treatments and then having frequent study visits and just seeing how your subjects are doing. Because what the most important thing they’re going to tell you is what is happening clinically.
So the EEG is going to be really interesting and really important, but until that’s sort of optimized, all the same as others have said, those clinical assessments, mood, memory, testing the organ that you’re targeting. So for me, hippocampus, we do a lot of memory stuff, we do mood stuff. It's different than if you’re doing other parts of the brain. So region specific clinical assessment, and then overall clinical assessment.
LENNART VERHAGEN: Bin He, you have a lot of experience in translating across different study models, both animals and humans. Are there best practices that we should all adopt?
BIN HE: Many have already been covered. I just want to share my own observation in our work usually is we try to be very careful in animal studies, like for example in the rodent model, we try and use (inaudible) pascal very low intensity focused ultrasound. But then in the histology assessment in these same animals we use eight times higher intensity, trying to make sure that focused ultrasound still is safe.
But by the way our recent results suggest that even up to eight times there is no histological structure damage observed. Going back to humans, and our practice is once again, because we are mainly studying healthy human subjects for neuroscience research, so we try to use as little as possible intensity, and while at the same time we try to have a very close follow-up with subject to make sure if there is any abnormality of the behavior, and also in subsequent after their TFAS and session, and if there will be anything that comes up. That is our experience so far.
LENNART VERHAGEN: Looking forward, what should we be doing? And maybe Til I’ll start with you. How can we make a prospective safety assessment?
TIL BERGMANN: As we’ve already discussed, I think it needs to be hypothesis driven, which means that we have to assess systematically the expected effect in detail, map out the secondary effects, and gather this information in a consortium to make sure that we get sufficient data a priori, instead of posthoc trying to figure out what might have happened.
LENNART VERHAGEN: Maybe while we’re rounding off I’m going to ask everybody to answer the one thing you feel we should be doing to tackle physiological or clinical safety? So you only get one sentence. Can I get started with Ellen? What is the one thing that we should be doing?
ELLEN BUBRICK: We have to keep studying it. If we don’t we won’t know. We have to be able to take risks to help our patients.
NOAH PHILIP: I am going to go with excitement, but I’m also going to go with very frequent sampling after folks get ultrasound.
MIRIAM KLEIN-FLUGGE: I think I would side with Til here. I think we need to standardize our reporting. Sharing the protocols that we use, the screening forms, the debrief forms and whatnot, especially when we have similar targets that are of interest.
TIL BERGMANN: I think (inaudible) research is the threshold for effective neuromodulation in terms of pressure - then we can really think about okay, what is on target, what is off target, a priori.
BIN HE: My suggestion is we should established optimized focused ultrasound with EEG simultaneous protocol so eventually we can take focused ultrasound neuromodulation to eventually a closed loop neuromodulation modality. Something that the DBS build has gone through for 30 years. But we’re just starting from. That’s my proposal.
LENNART VERHAGEN: I wish I could continue chatting with you for hours more, but I have to end. Let’s call this panel discussion to a close. Thank you all for your contributions.
KIM BUTTS PAULY: I think we are going to wrap up the session here. I just want to put a summary up of where we are. We had these questions. I think we’ve done a really good job of answering the questions, and that we have a lot of interesting stuff tomorrow we’ve got the regulatory pathway for FUS neuromodulation as a treatment, experimental planning and design, we’re going to get into more rigorous methods and explain those. And then optimizing target engagement and parameter space and effects. So I’m pretty excited about tomorrow. And Lennart, is there anything you want to say to round out the session?
LENNART VERHAGEN: I’ll look back at today. I’ve been in this field only for a few years, but I’ve learned so much, especially from this session. We had some fantastic cases on how strong the effects of neuromodulation can be in animals, but we’ve also seen that we’ve turned the corner with convincing evidence for human neuromodulation, we’ve been doing this for a few years now, and I think the field now is really convinced that we’re not just going on a rabbit chase, but there is something important and there is something real here, all sorts of future applications in mental health.
We’ve heard from the panel discussion on safety that this is not simply binary, there is a whole gradient on how we should be considering safety. There are very conservative, nearly arbitrary limits, and there have been fantastic discussions of how you can explore that further.
And also, here in the panel discussions on physiological safety, people were arguing for systematic mapping, to really start exploring a therapeutic window with very careful assessments, for example electrophysiological monitoring afterwards. That looks like a very bright future. I’m super excited about it. Thank you all for sharing, thank you all for your attendance. I’m looking forward to seeing you tomorrow.