A New Sensorized Approach Based on a DeepLabCut Model and IR Thermography for Characterizing the Thermal Profile in Knees During Exercise.

Abstract

The knee is one of the joints most vulnerable to disease and injury, particularly in athletes and older adults. Surface temperature monitoring provides insights into the health of the analysed area, supporting early diagnosis and monitoring of conditions such as osteoarthritis and tendon injuries. This study presents an innovative approach that combines infrared thermography techniques with a Resnet 152 (DeepLabCut based) to detect and monitor temperature variations across specific knee regions during repeated sit-to-stand exercises. Thermal profiles are then analysed in relation to weight distribution data collected using a Wii Balance Board during the exercise. DeepLabCut was used to automate the selection of the region of interest (ROI) for temperature assessments, improving data accuracy compared to traditional time-consuming semi-automatic methods. This integrative approach enables precise and marker-free measurements, offering clinically relevant data that can aid in the diagnosis of knee pathologies, evaluation of the rehabilitation progress, and assessment of treatment effectiveness. The results emphasize the potential of combining thermography with DeepLabCut-driven data analysis to develop accessible, non-invasive tools for joint health monitoring or preventive diagnostics of pathologies.