*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Lahiru Wimalasena
BME PhD Proposal Presentation
Date:2022-05-04
Time: 1PM - 2:30PM EST
Location / Meeting Link: https://emory.zoom.us/j/99006419044 (Meeting ID: 990 0641 9044)
Committee Members:
Chethan Pandarinath, Ph.D. (Advisor) Nicholas Au Yong, M.D., Ph.D. Lena Ting, Ph.D. Samuel Sober, Ph.D. Gordon Berman, Ph.D. Lee Miller, Ph.D. (Northwestern University)
Title: Investigating muscle pattern-generating mechanisms of the motor cortex and spinal cord using deep learning-based dynamical systems models
Abstract: There are approximately 17,000 new spinal cord injury cases every year in the US, representing only a small fraction of a growing patient population with unmet needs to restore motor function. A promising approach is to develop brain-machine interfaces (BMIs) that can restore function by directly stimulating muscles based on recorded signals from intact motor-related brain areas (e.g., motor cortex) – bypassing the malfunctioning part of the motor circuit (e.g., spinal cord injury). A long-term goal is to develop naturalistic BMIs, devices that can generate time-varying muscle activations with the same level of precision achieved by the motor system. In order to develop more naturalistic BMIs, there is a need for accurate descriptions of the motor system’s pattern-generating mechanisms that produce the moment-by-moment variations in multi-muscle activations to generate single-trial behavioral variability. Despite advancements in recording technologies, a challenging problem still remains of overcoming the limitations of noisy electrophysiological recordings, preventing accurate descriptions of the motor system on single trials. Recent advancements at the intersection of deep learning and computational neuroscience have resulted in novel dynamical systems modeling tools that de-noise cortical recordings to provide high-fidelity, moment-by-moment descriptions of neural population activity. To this date, however, no such tools exist for estimating multi-muscle activations, limiting our ability to study the muscle pattern-generating mechanisms of the motor system. This proposal aims to extend deep learning-based dynamical systems modeling tools to generate high-fidelity, moment-by-moment estimates of multi-muscle activations from EMG recordings (Aim 1). Using this class of dynamical systems modeling methods, I will investigate the pattern-generating properties of motor cortical neural populations (Aim 2) and spinal interneuron populations (Aim 3) and study their relationship to multi-muscle activation signals. This work will develop novel techniques for precise estimation of muscle activation, robustly test current hypotheses of motor cortical communication with muscles, and pioneer single-trial population-level analyses of spinal interneuron activity.
