M.Sc. Tezi Görüntüleme

      Pre-diagnosis varies depending on the experience of the physician. Therefore, patients are subjected to a lot of test required or unnecessary and it strains the patient in terms of finance and psychology. The most accurate diagnosis will be placed as a result of certain medical tests, but the success of pre-diagnosis accelerates the process of diagnosing. Hence pre-diagnosis system with low-cost, automatically, not changeable from one physician to another is needed. For this purpose in this study automatic pre-diagnosis program is developed by diagnosing the disease from face photos that is independent from sex, age and facial expressions.

      This study consists of four parts: creating databases, pre-processing, feature extraction and classification. Training database includes five image classes these are Acromegaly, Down syndrome, Lupus, Normal and other disease or non-face. Various pre-processing techniques applied on the training and test database images in the pre-processing stage, so images turned into same format and system performance improved. In the feature extraction, holistic approaches is studied comparatively. Basic approaches are used to classify the input image. In the last stage pre-diagnosis performance increased by applying the Borda Count method to the classifier results.

      Our proposed pre-diagnosis system shows %95 recognition performance.

      Key Words: Acromegaly, Down syndrome, Lupus, Feature Extraction, Local Patterns, Subspace-based methods, Classification, Diagnostic Tests, Borda Count