Mitosis Detection in Breast Cancer Histopathology Images Using Statistical, Color and Shape-Based Features

This paper presents an automated technique for mitosis detection in breast cancer histopathology images. Mitosis detection is the first step towards mitotic cell counting which is one of the metrics used for grading of breast cancer. A number of automated techniques for mitosis detection have been proposed in literature wherein different sets of features have been used such as textural, morphological, and statistical. The proposed scheme uses a novel combination of statistical, shape, and color-based features. Support Vector Machine (SVM) has been used to classify the candidate cells into mitotic and non-mitotic cells. The experiments on publicly available MITOS dataset show that the proposed technique outperforms the existing techniques by achieving precision, recall, and F-measure of 0.80, 0.90, and 0.85, respectively.