Automated Extraction of Anatomical Regions of Interest in Clinical Thermal Infrared Images of Hands

This paper presents a method for automated extraction of MCP joints in infrared images of hands. Based on averaged anatomical structure and landmarks of human hands, three different models are proposed to estimate the regions of MCP joints of human hand: rectangle model, ellipse model and circle model. In addition, to evaluate the accuracy and validity of the three models, we propose a new way to obtain ground truth images using a small circle paper board while collecting infrared images. According to the result and analysis, we get the conclusion that circle model has the best performance among the three.