Study on diagnosis of gastric tumors using feature parameter

At present, pathologists diagnose gastric tumor subjectively by atypism, pleomorphism and so on. Therefore the diagnosis are different by each of pathologists. In this paper, we propose a method to find differences among benign and malignant tumors, and gastric adenoma for the purpose of helping pathologists diagnose gastric tumors objectively. We extract nucleus and cytoplasm region from tissue images, and calculate feature parameters of them by image processing. By evaluating distribution of feature parameters from discriminant analysis, we find out the objective feature parameters for diagnosis.