NIMH MEG Core Facility

National Institute Of Mental Health

Investigators

Linked publications & trials

Abstract

Hardware Development: The MEG Core Facility has continued its development of a novel optically pumped magnetometer (OPM) system. With a combination of IRP and Brain Initiative funding (ZIA MH-0002975) we have acquired 64 sensors, with which are currently building a 56 channel array and three reference sensors (with additional spare sensors for redundancy). Array design work is in collaboration with George Dold and the Section on Instrumentation. During testing of our prototype 16 channel device, we observed inaccuracies in field measurements due to cross-axis projection error, or CAPE. CAPE is a phenomenon whereby changing ambient fields transverse to the measurement axis change the gain of the sensor and the orientation of the sensing axis. In order to mitigate this, we developed a technique we call cross-axis dynamic field compensation, which uses a fixed reference array to compensate for fluctuating ambient uniform magnetic fields; a manuscript describing our method has been accepted for publication at NeuroImage. Until we developed a compensation method to account for inaccuracies due to CAPE, precise calibration of our device was not possible. However, while testing the calibrator we noted several issues that prompted a redesign. With the new calibrator, which will contain an array of field generating coils, we will be able to more accurately model the field produced by the coils. In addition, the configuration of the coils will allow us to calibrate our sensor array more accurately. We expect to finish the construction of the new calibrator early in FY2023. In the interim, we have carried out pilot studies in healthy volunteers to determine the impact of our dynamic field compensation method on measurements of evoked responses to somatosensory stimuli, and data analysis is ongoing. All experiments involving human subjects occur under clinical protocol NCT04950309. Finally, during the FY2022 period, we also carried out comprehensive simulations of our 56 channel array, to demonstrate the utility for imaging and distinguishing simultaneous brain sources. Software Development: A variety of software for data analysis is maintained and supported by the Core. These include proprietary CTF code, beamformer source reconstruction software (the SAM suite) written in-house, and MNE-python. We also provide support for non-MEG specific software packages that these programs interface with, such as FreeSurfer and AFNI. In addition, the MEG Core Facility frequently writes custom scripts to integrate stimulus and response data with the MEG dataset. To supplement the SAM software suite, the MEG Core Facility has continued developing a modularized set of python scripts to automate all states of MEG data pre-processing and analysis. Due to the modular nature of these scripts, they can rapidly be customized for new studies. The MEG Core Facility has been implementing these scripts using data acquired as part of the NIMH research volunteer study (NCT03304665) and is currently working with individual Core Facility users to automate their analyses. These scripts are also being expanded to be deployed broadly across the MEG community as part of the ENIGMA project, as detailed below. The MEG Core Facility continues to assist investigators in setting up MEG software and ensures that all software is available on shared resources (the NIH High Performance Computing (HPC) center). We are continuing to develop a real-time neurofeedback task using source-localized MEG to modulate activity in the amygdala, in collaboration with Dr. Carlos Zarate's group. While MEG neurofeedback has been performed before, the use of source-localized neurofeedback is unique. It is much more complicated, however, requiring real-time beamforming and integration of the acquisition, reconstruction, and stimulus presentation software systems. Education and Training: One-on-one training and support are provided upon request, and accounts for a significant portion of the scientific staffs time. While we have not held in-person courses in the MEG Core Facility in FY21 due to the ongoing COVID-19 pandemic, we have expanded our online offerings. We have a once-weekly seminar that follows a rotating topic schedule between traditional journal club presentations, a machine learning in MEG special interest group, presentations of recent results by NIH investigators or extramural scientists, and tutorials. We have also hosted external software developers to provide demonstrations of new tools to the NIH MEG community. Support of the Larger MEG Community: Due to the ongoing COVID-19 pandemic, we have temporarily suspended the MEG North America meeting. While hosting a large MEG Hackathon gathering (such as the one we held in 2019) was not feasible, we did host a local gathering for NIH investigators. As soon as it is possible to run a meeting on the NIH campus with approximately 100 attendees, we intend to resume the meeting. Our intention is to avoid overlap with the BioMag meeting, which will be held in the fall of 2022. While we would love to hold an MEG North America meeting in the fall of 2023, this may not be an achievable goal. As part of the NIMH protocol Recruitment and Characterization of Healthy Research Volunteers for NIMH Intramural studies (NCT03304665) we collected 68 MEG recordings from healthy individuals on a comprehensive battery of cognitive tasks. These scans were part of a broader study collecting clinical, cognitive, and MRI data. In collaboration with the NIMH Data Science and Sharing Team, we converted all MEG data, as well as MRI, clinical, and behavioral data to the BIDS standard format, and shared the full dataset on the NIH-supported OpenNeuro repository. We have a data descriptor for the dataset accepted to Scientific Data. Our largest initiative in this domain has been the formation of the ENIGMA MEG working group. ENIGMA is a worldwide consortium of scientists in the domains of imaging and genomics. ENIGMA groups use meta-analysis or mega-analysis to understand how genotypes and neuroimaging phenotypes vary in health and a variety of neuropsychiatric disorders. We have recruited approximately 35 international MEG scientists to our working group. Currently, we have been collecting data from investigators willing to share raw anonymized data for processing at the NIMH. We have also collected resting state MEG scans from repositories including the Human Connectome Project (HCP), MEGUK, CAMCAN, and OMEGA datasets. We are nearing completion of a processing pipeline for the data analysis, which we expect to distribute to working group members in FY23. The vast majority of our working group members will be able to share derivative subject-level data, allowing us to perform a more powerful mega-analysis as opposed to a meta-analysis.