Population-level encoding of self-motion in… (F31DC023786) – Project Grant to The Johns Hopkins University

Abstract: Balance disorders affect millions globally, substantially reducing quality of life and elevating fall risk, with 35% of americans over 65 experiencing vestibular dysfunction. A leading cause is loss of vestibular function, the sensory system responsible for detecting head motion and orienting the body during movement. While multichannel vestibular prostheses have been developed to restore this function in patients with bilateral vestibular loss, current devices provide only partial restoration. This likely reflects a fundamental GAP in our understanding of how populations of neurons in the vestibular nuclei collectively process motion signals during natural behavior, since current prosthesis designs are based on models of individual neurons rather than the population-level computations that may govern balance control. This project aims to close that GAP. Specific aim 1 will identify VN encoding strategies and dynamics during locomotion by developing prism (population recording identification system for motion), a data-driven processing framework combining poisson gaussian process latent variable models (p-gplv) with universal differential equations (udes). Prism is expected to characterize the statistical structure of VN population activity and recover candidate dynamical equations linking head kinematics to neural population responses. Specific aim 2 will determine the VN neural manifold in vestibular-normal rhesus macaques during locomotion. IT is hypothesized that VN population activity lies on a low-dimensional manifold whose geometry reflects the six degrees of freedom of head motion. Alternatively, manifold structure may be organized by the activity of functionally distinct VN neuron classes rather than by head kinematics. Manifold stability across locomotion conditions and robustness across individual animals will be evaluated using prism. Specific aim 3 will elucidate structural changes in the VN neural manifold following BVL by comparing manifold dynamics between vestibular-normal and BVL conditions to determine whether compensatory population-level mechanisms may emerge in the absence of peripheral vestibular input. High-density neural probes with 128 channels will record simultaneously from VN neurons during locomotion in both vestibular-normal and BVL rhesus macaques, with synchronized 6-degree-of- freedom head kinematics and video-recorded gait analysis. This population-level, data-driven approach is expected to establish a quantitative framework for VN computation during natural movement. The resulting mathematical models are expected to provide the first quantitative, population-level description of vestibular nuclei computation during natural movement. These findings may directly inform the design of biologically grounded prosthetic stimulation strategies.

Population-level encoding of self-motion in the vestibular nuclei: toward data-driven models for restoring balance

Award facts

Description

Quick answers

Who received award F31DC023786?

How much is award F31DC023786 worth?

See this award's full record