Date of Award

6-2026

Degree Name

MS in Mechanical Engineering

Department/Program

Mechanical Engineering

College

College of Engineering

Advisor

Eric Espinoza-Wade

Advisor Department

Mechanical Engineering

Advisor College

College of Engineering

Abstract

Congenital hemiparesis is a unilateral motor impairment stemming from brain injury in utero or early postnatally. Hemiparesis can be difficult to detect in early infancy with current clinical tools. Yet, early identification of motor asymmetry could play a key role in the design of effective early intervention and rehabilitation strategies. This thesis presents the development and validation of a video-based tool designed to extract infant limb movement data using DeepLabCut’s (DLC) machine learning network. The pipeline was developed using videos of typically developing (TD) infants and infants with Asymmetric Perinatal Brain Injury (APBI), all at or under 3 months corrected age. A protocol was developed to crop, label, and process video footage for DLC to extract two-dimensional (2D) data of left/right wrist, and left/right ankle positions throughout the videos. A data processing pipeline was developed to remove outliers from the (2D) limb position data and a comparison was made between three different outlier detection methods. Finally, five metrics were applied to measure asymmetry: 1) the trace of the covariance, 2) principal component analysis (PCA) metrics, 3) the convex hull area, 4) the total clustered variance, and 5) the resulting asymmetry indices. The tool developed has the potential to detect side-to-side motor differences or limb asymmetry in infants with APBI early in development, and therefore, it is clinically significant.

Download

Share

COinS