Segmentation Based Classification for Mitotic Cells Detection on Breast Histopathological Images

Abstract

Breast cancer grading is the standard clinical practice for the prognosis and diagnosis of breast cancer development. The Nottingham Histological Grading (NHG) system is widely used in the breast cancer grading. One of the main criteria in assessing breast cancer is mitotic counts as it reflects the speed of cell division in cancer cells. Detection of mitotic candidates could be performed by implementing image processing techniques. The accuracy of mitotic cells detection is dependent on the number of mitotic candidates. Thus, minimizing the number of mitotic candidates is a crucial step in optimizing accuracy. This study proposed a segmentation based classification method to minimize number of false positive in mitotic candidates. The results show that the proposed segmentation based classification method could provide a promising result by achieving an average effectiveness of 91.85% in minimizing the mitotic candidate number.