A Non-Mitosis Reduction Method using Semantic Descriptors for Breast Cancer Mitosis Detection Application
Lu Juei Min, Tan Xiao Jian, Khairul Shakir Ab Rahman, Teoh Leong Hoe, Quah Yi Hang, Wong Chung Yee, Yip Sook Yee, Wan Zuki Azman Wan Muhamad, Teoh Chai Ling · 2022
Based on the Nottingham Histopathology Grading system, mitosis count is one of the important criteria that contribute to the overall grade of breast cancer. Over the years, many automated mitosis detection methods have been proposed. Nonetheless, the ever-increasing demand for quality detection continues by seeking optimization in each stage in the detection pipeline. This paper aims to focus on the optimization of the non-mitosis cells reduction stage by proposing three semantic descriptors: solidity, eccentricity, and area to eliminate the non-mitosis cells in breast histopathology images. The proposed method consists of three stages: (1) color normalization, (2) nucleus segmentation, and (3) non-mitosis reduction and its performance was evaluated using 40 histopathology images. The proposed three semantic descriptors were found to be useful and effective in reducing non-mitosis cells, achieving 96.18% (with standard deviation tabulated at ±1.6374%) across the dataset.