Cancer diagnosis using automatic mitotic cell detection and segmentation in histopathological images

Cancer is a disease characterized by abnormal cell growth in the human body. Cancer is evaluated by histopathological examination, which is important for further treatment planning. The tubule formation, mitotic cell count and nuclear pleomorphism are three parameter used for cancer grading. Mitotic cell (MC) count is one of important factor in cancer diagnosis from histopathological images. MC detection is very challenging task in cancer diagnosis because mitotic cell are small objects with a large variety of shapes. The aim is to evaluate performances of SVM (Support Vector Machine) classifier and Bayesian classifier in cancer diagnosis. This proposed work consists of three modules: 1) Pre-processing, 2) MC detection and segmentation, and 3) MC classification. MC detection and segmentation are performed by Bayesian modeling and local region threshold method. The segmented mitotic cell is classified by both SVM classifier and Bayesian classifier and their performance is evaluated.