Automatic Lesion Segmentation of Breast Ultrasound Image: An approach towards full automation

Abstract

Breast cancer is the most common cause of death among patients. One of the main reasons is that the lesion is not identified in proper time to seek medical facility. The situation in Bangladesh is alarming as there is a huge female population in the rural areas who don’t have proper medical access to detect the early stage of breast cancer. We have tried our utmost to contribute to the development of a Computer Aided Diagnosis (CAD) system that will detect the tumorous lesion in the BUS (breast ultrasound) image automatically. The algorithm will be able to come up with a Region of Interest (ROI) which eliminates human intervention in this phase. ROI generation phase consists of Horizontal cut based on entropy information and calculation of directional gradient from automatically generated seed point. Finally, we segmented our ROI image by adopting two approaches; one using watershed segmentation technique and other using entropy filtering. The thesis concludes with the comparison among the two segmentation methods.