Cancer Computing: The early detection of Breast Cancer

Abstract

Nowadays, breast cancer still the most common human death, however women are the most exposed, thus, its cause remains unknown. In order to reduce the associated morbidity and mortality, it would be necessary to detect this illness on its early stage. Several systems have been investigated to detect and diagnosis breast cancer in order to help radiologists in the screening step. CADe and CADx are the well known systems used to achieve this task. However, various steps are needed to obtain accurate results which consist in preprocessing, detection of abnormalities, classifying these latter as malignant or benign. After mammographic acquisition, some irrelevant elements must be removed from the mammogram such as Pectoral muscle, radiopaque artifacts, etc. For that reason we proposed, a novel method based on the similarity between intensities to delineate the pectoral muscle boundary using measure features of semantic similarity between words in Natural Language Process (NLP) and Information Retrieval (IR) elds. Also, the proposed method was exploited to classify the mammogram, based on its intensity, as normal or abnormal. The obtained results were promising since our approach gives an e ective pectoral muscle edge and extracts region with high intensity which improve the accuracy of our proposed system IOBI.