Date of Award

12-2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Biomedical Engineering and Sciences

First Advisor

Pengfei Dong

Second Advisor

James R. Brenner

Third Advisor

Peshala Priyadarshana Thibbotuwawa Gamage

Fourth Advisor

Mehmet Kaya

Abstract

This thesis evaluates whether accessible and affordable technologies can provide clinically significant, quantifiable gait analysis comparable to traditional marker-based motion capture systems. Modern high-end systems offer excellent accuracy but remain financially and logistically inaccessible to many clinicians and researchers. To address this gap, a framework was developed using markerless pose estimation using the model, MediaPipe, inertial measurement units (IMUs) processed through an Extended Kalman Filter (EKF), and surface electromyography (sEMG) to measure joint kinematics and neuromuscular activity during gait. An experiment was performed where data was collected from participants with both normal gait and pathologies including: Osgood–Schlatter disease, Plica Band Syndrome, and hip deficiency brought upon by osteoarthritis. Joint range-of-motion estimates generated by MediaPipe and IMU/EKF pipelines showed strong agreement, exceeding a Pearson correlation of 0.80. Distinct differences in knee kinematics were observed between injured and non-injured subjects, particularly during loading response and mid-swing. sEMG data showed a poor correlation between range of motion in the knee and neuromuscular amplitude compared to expected results such as quadriceps avoidance and hamstring facilitation in individuals with knee pathology. These results demonstrate that the a markerless approach can reliably characterize lower-limb motion and muscle activation at a fraction of the cost of professional systems. The findings support the development of accessible gait analysis methods that can enhance clinical diagnostics, rehabilitation monitoring, and broader research applications. Continued refinement of calibration, data standardization, and larger participant sampling will further strengthen the clinical viability of this model.

Download

Share

COinS