The Maturation of Functional Brain Networks: Insight into the Origins and Course of Mental Disorders
- Sponsored by:
- National Institute of Mental Health
Neural networks are a fundamental property of normal brain function, and dysregulated brain activity has been implicated in a wide array of mental disorders. In January 2011, NIMH convened a multi-disciplinary workshop of experts to discuss the opportunities and challenges of studying the maturation of neural networks in healthy and clinical populations. The workshop focused on the following topics in order to specify promising strategies for advancing the field:
Topic 1: New Approaches to Study Neural Network Maturation
Much of the work on neural networks in functional magnetic resonance imaging (fMRI) studies has come from analyses of resting-state fMRI. “Resting-state” refers to spontaneous brain activity in participants not explicitly engaged in a task. Scientists analyze how activity in one brain region is related to activity in other regions, as a way to understand brain “connectivity”. There has been active debate about the functional significance of resting-state brain activity, but the patterns of activity are nevertheless robust and consistently observed across varying experimental settings. Many participants viewed it as a complement to data from task-based studies, and as a way to understand overall brain connectivity. Resting-state functional connectivity analysis has been especially useful in studying neural networks in infants and young children, who are not cognitively mature enough to participate in task-based imaging. In addition, as little as five minutes of scanning can provide sufficient data for resting-state functional connectivity analyses1—an advantage for participants for whom lying still for extended periods of time can be challenging.
Functional connectivity analyses are not limited to fMRI. Participants presented data on recent advances in electroencephalography and optical imaging. These methods measure brain activity by examining its electrical signals or by inferring how much energy the brain is using, respectively. A major advantage of these methods is that the interface is a portable cap worn by the participant, thus greatly improving the applicability for young children. However, participants noted the need for more than the development of new tools, per se, but also increasing the utility of existing tools (e.g., developing better optical imaging caps specifically for very young children).
Data gathered from these multiple modalities can provide promising insights into the complexities of neural networks and mental illness. For example, biomarkers of mental illness can be created by applying computational methods, such as support vector machine (SVM) analyses. SVM-based methods rank brain connectivity measures by how well they associate with pathology, and produces metrics that may be predictive of individual diagnostic status or treatment response.
Topic 2: Structure-Function Relationships over the Course of Development
Infancy through young adulthood is a time of striking changes in brain development. But to characterize these changes as merely “bigger” (e.g. brain volume) or “more” (e.g. myelination) oversimplifies the process. Instead, regional changes in brain maturation can shift from subtle to dramatic at different developmental periods, and are influenced by a complex array of genetic and environmental factors. It is important not only to consider the endpoints of brain development, but also the path itself. The journey and not just the destination will provide critical insights into the etiology and pathophysiology of mental illness. Participants also discussed the challenges of scaling metrics from an adult brain to a developing one, as a child’s brain is not simply a smaller version of the adult brain. The characteristics and functions of a neural network can be different at varying developmental stages.
Discussion also centered on understanding the cellular mechanisms that underlie the data obtained from large-scale structural and functional network analyses. For example, synaptic pruning is often an inferred mechanism for the observed changes in structural and functional connectivity during adolescence. However, the causal links between these events have yet to be determined. Parallel studies in humans and animals would be an effective way of answering these fundamental questions. Approaches such as in vivo imaging of animal brains, with subsequent fine-grain anatomical studies, are another way to bridge levels of analyses. An additional area of opportunity is to increase integration of computational neuroscience to more accurately model brain networks during development.
Topic 3: Maturation of Neural Networks in Typical and Atypical Development
Attention then turned to our current knowledge of neural network trajectories in mental illness. Participants presented data on functional connectivity in attention deficit hyperactivity disorder (ADHD) and Tourette syndrome. Previous longitudinal studies in typically developing children have shown a consistent “local-to-distributed” maturation pattern—with age, there is decreased connectivity in spatially close brain regions and increased connectivity in more spatially distant areas2,3. Interestingly, connectivity is “immature” in ADHD and Tourette syndrome. Neural networks in children with ADHD and Tourette are more similar to those in younger children than to those in age-matched children. A presentation on motor network connectivity in autism spectrum disorder showed that bias to proprioceptive information and increased motor cortex white matter predicted social and motor skill impairment. These findings indicate that studying a domain-specific network can be an effective approach to understanding the mechanisms that underlie the broader deficits of a disorder. Accordingly, a continuing goal will be to understand how cognitive and affective functions emerge from network activity, and how aberrant network dynamics relate to disorder symptomatology.
Participants discussed the best age ranges to study for understanding the development of functional neural networks. While some argued that starting as early as possible was best, other emphasized the utility of studying older age groups, especially if the age range was within a developmentally sensitive period (e.g., adolescence). However, all participants acknowledged that starting early would help disentangle the roots of disorder from secondary or compensatory effects. Furthermore, it is important to gather data on participants prior to disease onset, in order to identify biomarkers which would be vital to the development of pre-emptive interventions.
Topic 4: Opportunities and Challenges in Neural Network Studies Using a Prospective, Longitudinal Design
The session focused on improving study designs in clinical populations—a difficult task given the heterogeneity, co-morbidity, and changing medication status in patients with mental illness. Participants felt a shift was needed away from the use of highly narrow criteria for subject inclusion; instead, emphasize subject characterization. A dimensional approach of deep phenotyping was viewed as a more tractable method of revealing the origins and developmental trajectories of mental illness.
Longitudinal studies allow characterization of individual growth curves and can capture abrupt changes that occur at sensitive periods in development. However, these studies are also time-consuming and more expensive than cross-sectional studies are. A mixed design is a promising way to optimize the pros and cons of each approach. For example, a study might consist of cross-sectional age bands, with longitudinal assessments within each band.
Lastly, workshop participants agreed that a reference map of connectivity in typical development was needed. A large database could be leveraged by other studies and would facilitate the development of more accurate and predictive computational models of neural network maturation. Accordingly, data sharing is critical to advancing the field. Changing the culture from viewing data as proprietary to one more of community science remains an ongoing goal.
In summary, this workshop highlighted recent advances in methodologies and knowledge of neural network maturation. Participants identified a number of research gaps and opportunities that included:
- Continued development of better analytical tools for scaling developmental data and for interpreting multi-modal data. Emphasis should also be placed on improving the usability and accessibility of tools in young children;
- Integration of animal studies and computational models with human studies to gain insights into the mechanisms of dynamic neural networks over development;
- Tracking the developmental trajectory as early as possible, with a focus on individual, instead of group data, in order to develop personalized biomarkers;
- Deep phenotyping of study participants using dimensional approaches, rather than trying to obtain “pure” clinical populations by imposing overly narrow criteria;
- Increased data and tool sharing to facilitate both hypothesis-driven and discovery-based research on neural network maturation; and,
- Comprehensive “mapping” of normative development of structural and functional connectivity in order to accurately interpret data gathered on clinical populations.
For more information, please contact Kathleen Anderson, Ph.D., firstname.lastname@example.org, 301-443-5944.
- Van Dijk, K.R., et al., Intrinsic functional connectivity as a tool for human connectomics: theory, properties, and optimization. J Neurophysiol, 2010. 103(1): p. 297-321.
- Fair, D.A., et al., Functional brain networks develop from a “local to distributed” organization. PLoS Comput Biol, 2009. 5(5): p. e1000381.
- Supekar, K., M. Musen, and V. Menon, Development of large-scale functional brain networks in children. PLoS Biol, 2009. 7(7): p. e1000157.