Segmentation of Tumour Regions for Tubule Formation Assessment on Breast Cancer Histopathology Images

Abstract

Breast cancer is the second leading cause of cancer death in women worldwide. In Malaysia, one out of 20 women will develop breast cancer once in their lifetime. Nottingham Histological Grading (NHG) system is widely accepted worldwide as the gold standard in providing the overall grade to breast cancer. Tubule formation is an important breast cancer feature used in the NHG system. Assessment of tubule formation requires pathologist to identify tumour regions. However, colour variation on breast histopathology images could provide nonstandard imaging and influence tumour regions detection. Manual identification of tumour regions using microscope may also vary between pathologists. Thus, automatic segmentation is crucial to detect tumour regions. By eliminating the irrelevant regions on breast histopathology images using automatic segmentation method, pathologist can focus on the tumour regions for further investigation. This study proposed a simple segmentation using global thresholding method to segment tumour regions on breast histopathology images. Partial contrast stretching and median filter were applied as pre-processing to provide better image quality for further segmentation processing. Post-processing using morphological operation and boundary padding were applied to enhance the segmentation results. The proposed method was able to segment the tumour regions on breast histopathology images and obtained accuracy, sensitivity and specificity of 84.22%, 85.83% and 84.89% respectively.