Advancing knee injury prevention and anomaly detection in rugby players through automated processing of infrared thermography: A novel biothermodynamics approach

Infrared thermography is increasingly used in medical and sports science as a non-invasive technique for monitoring physiological processes. Despite its potential, current practices rely heavily on manually defined regions of interest (ROIs), which introduce variability, reduce repeatability, and hinder large-scale applications. To overcome these limitations, there is a growing need for automated, standardized approaches that ensure consistent and objective thermal analysis. This study proposes a novel methodology that integrates visual and infrared imaging to automatically detect and segment ROIs using the YOLO (You Only Look Once) object detection model. By combining these modalities within a unified framework, the approach reduces human error, accelerates analysis, and enhances the precision and reliability of data extraction. A pilot case study involving rugby players was conducted to validate the method under realistic conditions. The system achieved high segmentation accuracy, with a Dice coefficient of 0.99, and successfully identified thermal patterns such as hyperthermic spots, asymmetries, and potential varicose veins. The methodology also offers valuable insights into the interplay between biothermodynamics, environmental factors, and heat transfer mechanisms that shape thermal imprints. These findings demonstrate the feasibility and advantages of using automated, multimodal thermal imaging for both qualitative and quantitative assessments. While this work is exploratory and does not constitute clinical validation, it lays the groundwork for future research aimed at enhancing diagnostic accuracy and standardization in sports medicine and broader medical diagnostics.