Establishing Standards for Data and Metadata from Wearable Devices
Presenter
Gregory Farber, Ph.D.
Division of Data Science and Technology
Goal
Wearable technology shows promise in providing information that can be used to create biomarkers of various diagnostic groups that have relevance to mental illness. However, there are barriers related to the different data formats used in existing devices that make data aggregation difficult. The purpose of this concept is to support the establishment of standards for data, as well as metadata and related coordination activities, that will allow researchers to easily access data from wearable devices and integrate those data for subsequent data analysis.
Rationale
Data from wearable devices is increasingly being used by NIMH researchers for a variety of purposes. From the data management and sharing plans submitted for the October 2024 Council round, we see that a significant number of applications had plans to deposit data from wearable devices in a data archive. The burgeoning interest in this sort of data by the mental health research community can be easily explained. Wearable devices give researchers the real opportunity to monitor behavior in a quantitative way.
There are currently no meaningful standards for either the data from wearable devices or the metadata related to data collection parameters from wearable devices. Without standards, data from these devices cannot easily be reused or analyzed by the research community. A recent meeting outlined the current challenges for the field (https://escholarship.org/uc/ucla_depression_grandchallenge_digsen ). The number of investigators planning to collect such data for NIMH-funded research combined with the identified challenges suggests that NIMH should act now to help the field generate standards for data and metadata from wearable devices.
Wearable devices have the potential to provide information that may turn out to be very useful in creating biomarkers for groups of people who have a mental illness. However, wearable devices and cell phones are created by a number of different manufacturers each of whom deliver analyzed data (for example, about sleep quality) to consumers. In some cases, the raw data are also made available to the consumer, but the data format for either raw or derived data streams differs from manufacturer to manufacturer. This makes it difficult or impossible to link data sets coming from different devices.
Other areas of data collection have had similar issues and have found solutions to those problems. For example, in the magnetic resonance imaging field, each of the device manufacturers collect data using different protocols and have different internal data formats compared to the other manufacturers. However, that community has developed the Digital Imaging and Communications in Medicine (DICOM) data standard that allows each manufacturer to export data using the same standard. In turn, this has allowed researchers to develop analysis tools and pipelines that begin with data in DICOM format. Such standardization has been essential for data analysis in the field. DICOM is not the only standard that is useful in the magnetic resonance space. Others such as the Neuroimaging Informatics Technology Initiative (NIfTI) and its derivatives have been very useful to the research community. NIMH played a key role in the development of NIfTI.
The goal of this concept is to develop standards and data formats like DICOM and NIfTI that will promote re-use of data from wearable devices and cell phones. Grant recipients will work with the research community, device manufacturers, and potentially other government agencies to develop these standards. The awardee will not only work with those communities to develop the initial standards, but they will provide services related to implementing the standards and rectifying any shortcomings once the first version of standards has been created and released. Given the pace of change in this field, those services may include ongoing development of new versions of the initial standards that will be backward compatible with the initial version of the standard.