Design and Development of Algorithms for Automatic Breast Cancer Detection

Abstract

Breast cancer is the most common type of cancer in women worldwide, and the leading cause of death from cancer in women, especially those between 40 and 55 years of age. The early detection and accurate diagnosis of breast cancer is of utmost importance in providing effective treatment in order to increase survival rates. Mammography is the effective, economic and widely used imaging modality for breast cancer detection. Masses are the most common indicators of breast cancer which are predominant on mammograms. Two types of techniques such as screening and diagnostic mammography are routinely performed by radiologists who interpret mammograms by visual inspection. However, manual inspection is a tiring and tedious task prone to errors. Some retrospective studies have revealed that around 10-15% of breast cancer cases are missed by radiologists. In this way, the search for Computer-Aided Detection/Diagnosis (CAD) techniques has been encouraged for effective early detection of breast cancer at most treatable stage followed by successful diagnosis reducing error rate and improving decisions of the radiologists. The fundamental aim of this research work is to design and develop image processing and machine learning algorithms to investigate the potential of mammographic texture to identify breast abnormalities and characterize them as either benign mass or malignant tumors. Contributions are made in the different phases, including: (1) Preprocessing, (2) Detection of suspicious abnormal lesions, (3) Characterization of suspicious lesions and their classification, and (4) Breast cancer diagnosis using ipsilateral views. The goal of this research is to increase the diagnostic accuracy of image processing and machine learning algorithms for optimum classification between malignant and benign abnormalities in digital mammograms by reducing the number of misclassified cancers. All the developed automatic algorithms are thoroughly evaluated using a database of full field digital mammogram images obtained from Tata Memorial