ePoster

Coordinated cortico-cerebellar neural dynamics underlying neuroprosthetic learning

Aamir Abbasi,Andrew Fealy,Nathan Danielsen,Tanuj Gulati
COSYNE 2022(2022)
Lisbon, Portugal
Presented: Mar 19, 2022

Conference

COSYNE 2022

Lisbon, Portugal

Resources

Authors & Affiliations

Aamir Abbasi,Andrew Fealy,Nathan Danielsen,Tanuj Gulati

Abstract

Brain-machine interfaces (BMIs) or neuroprosthetics allow neural control over assistive devices. They also provide an important framework for studying neural plasticity. It has been suggested that learning is essential for robust neuroprosthetic control. However, little is known about neural processes in subcortical regions that support learning a neuroprosthetic skill. Recently, cortico-striatal interactions during neuroprosthetic learning have been explored5 but what are the emergent cerebellar dynamics with cortical neuroprosthetic control is unknown. We performed simultaneous electrophysiologic recordings in the motor cortex (M1) and the cerebellum (Cb) of rats while they used M1 activity for direct neuroprosthetic control on the angular velocity of a feeding tube. We analyzed how the activity of M1 ‘direct’ neurons controlling the tube, as well as other recorded ‘indirect’ neurons in M1 and Cb changed while learning the neuroprosthetic task. We also analyzed band-limited activity in local-field potentials (LFPs) in both regions. Furthermore, we performed optogenetic silencing of Cb, by injecting red-shifted halorhodopsin (Jaws), while rats performed the neuroprosthetic task and analyzed how cerebellar silencing affected the M1 activity and neuroprosthetic learning. We found that learning successful BMI control was associated with robust modulation of ‘direct’ neurons in M1 along with robust ‘indirect’ modulation in M1 and Cb. Furthermore, we observed the emergence of task-related 3-6 Hz synchronous activity in cortico-cerebellar LFPs. We found that these trends emerged late in learning once skillful BMI control was learned. Furthermore, Cb inhibition led to poor performance and weak M1 activity, causally demonstrating the necessity of Cb for skillful BMI control. Our work has identified neural mechanisms in M1 and Cb which are associated with learning of a cortically controlled neuroprosthetic task. This underscores the importance of optimal engagement of neural learning mechanisms in an offsite motor region- the cerebellum, for successful learning of M1-controlled neuroprosthetic task.

Unique ID: cosyne-22/coordinated-corticocerebellar-neural-a6aaa460