Non-Mitosis Candidate Reduction Method for Breast Carcinoma Mitosis Detection Application

Lu, Juei Min (2022) Non-Mitosis Candidate Reduction Method for Breast Carcinoma Mitosis Detection Application. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Lu Juei Min.pdf Restricted to Registered users only Download (4MB)

Abstract

Based on the Nottingham Histopathology Grading (NHG) system, mitosis cell detection is one of the important criteria that contribute to the overall grade of breast cancer. This project aims to reduce the number of false positive candidates, which is non-mitosis cells after the nucleus segmentation process. The project consists of 3 steps: (1) color normalization to standardize the color of the RGB Hematoxylin and Eosin (H&E) stained input images (where the nuclei structures and cytoplasm are in blue and pink color, respectively), (2) nucleus segmentation would be implemented to segment the nucleus cells and lastly, (3) false positive candidates reduction to reduce the number of non-mitosis cells. However, the unbalanced input data into the classifier after the segmentation stage could impinge the overall classification results in mitosis cell detection. The proposed false positive candidates reduction method aims to reduce the number of non-mitosis cells. The overall effectiveness and average accuracy results of this project are 96.18% and 88.71% respectively.