FACIAL BLOCK ANALYSIS FOR MEDICAL APPLICATIONS by Shu Ting Thesis Supervisor: Assistant Professor, Bob Zhang Software Engineering Since the traditional diagnostic method of most diseases is a blood test, which is painful and invasive, it is necessary to develop a non-invasive method in order to appeal to the masses. According to Traditional Chinese Medicine theory, the facial image of a person can reflect the inner health status of that individual. Therefore, quantitative facial image analysis is an important way to detect disease. However, disease detection through facial image analysis is a challenging problem, because of the complex diversification of a facial image. How to detect disease (especially Diabetes Mellitus) based on facial block feature analysis will be presented in this thesis. Facial block color and texture feature extraction will be introduced followed by a description of our proposed methods: three methods at Diabetes Mellitus detection and one method at disease detection. The three Diabetes Mellitus detection methods are (1) using Dictionary Learning with the Sparse Representation Classifier based on facial color features; (2) based on facial texture features using Simplified Patch Ordering and Improved Patch Ordering; (3) through k-Nearest Neighbor and Support Vector Machine based on facial texture features. The disease detection method is based on facial texture features using k-Nearest Neighbor and Library of Support Vector Machine. According to the four methods, facial block feature analysis is proven to have good performance at disease detection.