A machine learning based framework for assisting pathologists in grading and counting of breast cancer cells

Abstract

Breast cancer normally occurs in the breast cells of both men and women, but is prominent in women. Computer aided detection increases the chance of early detection and diagnosis. This paper proposes a breast cancer detection method using Nuclear Atypia Scoring (NAS). The proposed cancer detection method works by converting each and every cancerous tissue into objects. Along with detecting the grade, proposed mechanism gives the count of the detected cells. This assists pathologists in identifying whether cells are cancerous or not along with the count of each type. Proposed model was evaluated on MITOS-ATYPIA-14 Challenge dataset. Accuracy of 0.89 and precision of 0.87 is obtained by the best method. Results indicate that the proposed machine learning technique has better performance as compared to existing methods and can aid pathologists in the detection process.

Keywords

• Breast cancer
• NAS scoring
• Machine learning
• Deep learning