Fully Automatic Computer-Aided Detection of Breast Cancer based on Genetic Algorithm Optimization

Breast cancer is the second disease causing death in women. The screening of breast cancer for early detection is an effective way to reduce mortality. Generally, mammography is the most effective imaging modality recommended by radiologists. Therefore, a high number of mammogram images have to be analyzed. Indeed, the expertise of radiologists and the quality of the images to be analyzed play a crucial role in the accuracy of the diagnosis. The use of CAD systems allow improving the accuracy of diagnosis by reducing the number of false-positive and the false-negative. In this paper, to assist radiologists to make the right decision by showing them the probably suspect area, a full automatic CADe is presented. The preprocessing step aims at delimiting the ROI by removing artifact and pectoral muscle using global thresholding method, morphological operators and seeded region growing method, followed by contrast enhancement of the ROI found in the first step using the technique named “adaptive local gray level transformation based on variable s-curve”. At last, the ROI is segmented using k-means clustering. Both contrast enhancement method and the segmentation method are performed by genetic algorithm to optimize the outcomes of each step.