Research And Implementation Of Cad On The Pathological Slices Of Gastric Adenocarcinoma

Computer-aided medical image processing and analysis is an important application in the field of image processing with medical image characteristics. It is absolutely helpful for clinical diagnosis and medical studies that computer-aided diagnosis system of cancerous tissues is introduced, especially in the case of lack of specialists. The present computer-aided diagnosis on cancer focuses mainly on the studies of cell smear. However, the cell image of pathological slices, which is most commonly found, is scarcely studied for the organizational complexity and difficulty in dividing the cells.The computer-aided diagnosis on the pathological slices of gastric adenocarcinoma is studied in this thesis. The complex image preprocessing of cell images, calculation of segmentation of nucleus outline, feature extraction and discriminant of nucleus are the main tasks.During the phase of preprocessing of cell images, based on the complexity of cell images of pathological slice from gastric cancer, the related preprocessing techniques are analyzed and the complex preprocessing technique are proposed, which is based on thresholding and aims at delimiting the effects of other tissues.During the phase of nucleus segmentation, the combinational algorithm, watershed transformation with GVF Snake model, is laid out. The initial contour of nucleus is achieved by watershed transformation, then the real contours are approaching by GVF Snake.In the phase of the feature extraction and discriminant of nucleus, the traditional method is discarded and the statistical Fisher is adopted to differentiate and analyze the large sample data. Finally, the Fisher discriminant function expression is established by iterative screening on characteristic parameters and solution of coefficients of characteristics parameters.According to the experiment, the improved computer-aided diagnosis system on the pathological slices of gastric adenocarcinoma possesses satisfactory cell segmentation capacity and cell classification capacity, which can play a role in pathological diagnosis and can improve the ability of junior residents in differentiate pathological section. Theoretically, the study broadens the scope of computer-aided diagnosis and establishes a solid foundation for similar studies in the future.