Breast cancer prevails to be one of the major cancer types in women worldwide. In the context of the prevention and early detection of breast cancer, it is necessary to conceive diagnostic aid systems to control tumor growth with very high precision and to give more effective treatments adapted to the pathological stage of the tumor. To achieve such a system, there is a need to firstly start with the pretreatment step which enhances the quality of boundaries and structures. Secondly, the segmentation step is needed to be performed on the pretreated images which consist of the pectoral muscle (PM) that is located in the upper corner of the mammography in the Medio-Lateral-Oblique (MLO) view. In this paper, we propose a novel approach to remove PM in MLO observations of mammograms. This approach is based on the ideas of clustering, region and edge. For implementation, experimentation and verification, the proposed technique has been tested on the digital mammography of Mini-MIAS database. DICE Coefficient and Structural Similarity measure have been used to find out the goodness of segmentation between the segmented regions and the ground truth. The proposed approach has proved to be effective and superior as compared to various existing techniques in the same context.
Khoulqi, Ichrak; Idrissi, Najlae; and Sarfraz, Muhammad
"Segmentation of Pectoral Muscle in Mammogram Images using K-means and Region Growing,"
Information Sciences Letters: Vol. 10
, Article 7.
Available at: https://digitalcommons.aaru.edu.jo/isl/vol10/iss1/7