An Algorithm for Supervised Segmentation of Infrared Foot Sole Images Based on a Bayesian Classification Model

Diabetes Mellitus (DM) continues to be a significant challenge for public health, with a considerable impact on global mortality rates. One of its most severe complications is diabetic foot syndrome, which, without proper treatment, can lead to ulceration and amputations. In this context, medical thermography offers a non-invasive approach to detecting thermal abnormalities in the skin, positioning it as a promising tool for the early detection of these complications. This paper presents a segmentation algorithm applied to thermographic images of the plantar region, developed using a Bayesian classification model. This model uses color features from the images to effectively segment the plantar region achieving a sensitivity of 0.7574. However, increasing the color samples could further enhance the model especially by variations in image acquisition conditions (e.g., lighting, sensor calibration). Future work will focus on expanding the dataset, adaptive color sampling, and alternative feature extraction methods to enhance robustness and segmentation sensitivity.