Single-Camera System for Hand Injury Assessment

Abstract

The hand plays a vital role in human functionality but is frequently vulnerable to injuries, particularly in occupational settings, with around 1 million workers visiting emergency rooms annually for hand-related incidents. Recovery assessment post-injury is complex and costly, often involving significant time and financial burdens on patients. While advanced surgical techniques are available, optimal recovery necessitates effective rehabilitation and functional monitoring, as neglecting follow-up can lead to complications that impair hand function. Traditional evaluation methods for hand injuries often rely on subjective assessments and lack precision, hindering rehabilitation progress. There is an increasing demand for automated hand function evaluation, which requires accurate measurement of hand movements. This paper introduces a cost-effective computer vision-based system using a single camera to assess hand recovery levels. Employing MediaPipe for joint motion analysis, the system calculates finger bending angles and evaluates motion smoothness. Additionally, the Jebsen-Taylor Hand Function Test (JTHFT) is incorporated, analyzed through the YOLOv8s object detection model to monitor task execution. This innovative approach aims to enhance the accuracy and accessibility of hand function assessments, ultimately supporting better recovery outcomes for patients with hand injuries.

Lab Members

Patrick Chan

Associate Professor, Vice Dean

Patrick Chan works on machine learning, deep learning, image processing, adversarial learning, and secure machine learning.