In recent years digestive colon diseases have increased dramatically. A category of common digestive diseases are Colitis. In medicine, Colitis refers to inflammation of the colon. Collagenous Colitis is a type of microscopic colitis which is characterized by chronic and intense diarrhea. While this change is normal in macroscopic view of the colon tissue, it is abnormal in the microscopic view. Routine tests like blood tests and parasite tests cannot be used for diagnosis. The only method of diagnosis is sampling of the colon during colonoscopy and measuring the thickness of collagen band by Pathologist which is obtained through microscope images. Thickness of collagen band in the colon tissue of a patient is more than 10 micrometers. This change of collagen thickness in sub epithelial may occur anywhere in colon tissue, so pathologists need to sample several areas of colon tissue and measure the thickness of the collagen band. Now this procedure is done manually, which leads to fatigue and loss of accuracy of pathologists. In this thesis, we take high resolution images from tissue samples in order to develop the algorithm which can extract collagen band from colon tissue and measure the thickness of the collagen band accurately. The proposed algorithm was designed in a way to make it resistant to the changes of environmental conditions, and acceptably flexible to segment colon tissue images. Due to the fact that the image segmentation of colon tissues needs to have high level quality, the first step is preprocessing. Then the color and texture features must be extracted from the image. After that the optimum features which perform better in the separation of target and nontarget image samples are selected through the Adaboost feature selector. In the next step accurate optimized classifier is used to separate collagen band from the colon tissue. Through several different methods in this research, MLP neural network is obtained as the most efficient way. In the last step, after completing segmentation by using appropriate algorithms, the thickness of collagen band is measured along the axis passing through its area. Performance of the proposed algorithm and accuracy of the results have been confirmed by a pathologist. Keywords: Collagenous colitis, Image segmentation, Feature extraction, Classification, MLP Neural network
