Toward an Enhanced Approach for Pectoral Muscle Segmentation in MLO View Mammograms

Breast cancer (BC) remains a predominant concern among women globally. Striving to prevent and identify BC in its initial phases, the development of computer aided-diagnosis (CAD)is imperative. These systems play a pivotal role in precisely controlling tumor growth and administering tailored treatments based on the tumor's pathological stage. The foundational step in creating such a system involves a crucial pretreatment phase aimed at enhancing the quality of image boundaries and structures. Subsequently, the segmentation step becomes essential, particularly in the context of Medio-Lateral-Oblique (MLO) view mammograms, where the images encompass the pectoral muscle (PM) situated in the upper corner. This paper introduces a novel approach for PM removal in MLO mammogram observations, leveraging region, and edge-based concepts. The proposed technique has been rigorously tested on digital mammography from the Mini-MIAS database, with the evaluation metrics of DICE Coefficient, Structural Similarity (SSIM) and Jaccard Similarty Index (JSI) providing insights into the segmentation quality against the ground truth. The results affirm the effectiveness and superiority of the proposed approach in comparison to several other methods within the identical field.