Analysis of Bimanual Motor Task Function in the Mouse Model After Ischemic Stroke Utilizing Deep-Learning Model DeepEthogram

Olivia S. Jung

Document Type

Article

Publication Date

2025

Keywords

JMG

Abstract

Ischemic strokes are the most common stroke type and result from blockages in neural blood vessels, restricting oxygen flow and causing an ischemic event. These strokes often lead to disabilities that impair motor function of stroke survivors, especially their ability to use both hands at the same time through bimanual coordination. DeepEthogram is a program that utilizes a deep-learning model, allowing for analysis of a bimanual motor task. We use the string-pull task, in our mouse models of ischemic stroke, to study mechanisms of neurological function that could enable post-stroke motor function to return to pre-stroke patterns. The goal of this project was to create a DeepEthogram model to quantify motor behaviors over post-stroke time points and determine any correlations between motor deficit with stroke volume. A success-failure DeepEthogram model was trained to 92% accuracy to automate distinguishing right and left paw successes and failures during the string-pull task for four behaviors: RSuccess, RFail, LSuccess, and Lfail. This model was applied to videos of 10 animals of equal sexes, over 56 days post-stroke. Analysis revealed that the motor deficit and recovery of the right (paretic) paw occurred as expected with RSuccess occurrences to be highest pre-stroke, lowest at day 3 in the acute phase of stroke, and recovery to pre-stroke levels at day 56 in the chronic phase. In parallel, the lowest occurrence of RFails was lowest pre-stroke, highest at day 3, and recovered to pre-stroke levels at day 56. The left (non-paretic) paw is more variable although day 56 and pre-stroke values were not statistically different, indicating recovery to pre-stroke patterns. Only one statistically significant difference between sexes were determined at day 21 for LSuccess (p=0.007). A subsequent analysis for motor deficit differences, by measuring changes in RFail occurrences at day 7 from pre-stroke baselines, by sex demonstrated a lack of significant differences (p=0.133) when the linear mixed effects model was applied. An analysis of stroke volume with immunohistochemistry revealed little correlation between stroke volume and motor deficit with R2=0.293 indicating a lack of effect of the stroke volume on motor deficit. There were no statistically significant differences (p=0.255) in the stroke areas between males and females for slices containing the stroke and stroke volume, even when the outlier SPT1 with a z-score of 3.53 is included, when the linear mixed effects model was applied with stroke area as a function of sex. Bimanual coordination is a significant feature of many daily tasks of humans and observing patterns of behavior in the string-pull task pre- and post-stroke in the mouse model could allow the discovery of novel mechanisms of motor function recovery to inform current neurohabilitative strategies.

Please contact the Joan Staats Library for information regarding this document.

COinS