Textural Approach for Mass Abnormality Segmentation in Mammographic Images

TL;DR

This paper explores a contour-based GLCM texture feature approach for segmenting breast masses in mammograms, tested across different tissue densities, showing promising results compared to edge-based methods.

Contribution

It introduces a contour-based GLCM texture analysis method for mass segmentation, expanding the application of GLCM beyond region-based approaches in mammography.

Findings

Promising segmentation results on challenging mammogram images.

Effective differentiation of tissue densities in mass segmentation.

Potential improvement over traditional edge-based segmentation methods.

Abstract

Mass abnormality segmentation is a vital step for the medical diagnostic process and is attracting more and more the interest of many research groups. Currently, most of the works achieved in this area have used the Gray Level Co-occurrence Matrix (GLCM) as texture features with a region-based approach. These features come in previous phase for segmentation stage or are using as inputs to classification stage. The work discussed in this paper attempts to experiment the GLCM method under a contour-based approach. Besides, we experiment the proposed approach on various tissues densities to bring more significant results. At this end, we explored some challenging breast images from BIRADS medical Data Base. Our first experimentations showed promising results with regard to the edges mass segmentation methods. This paper discusses first the main works achieved in this area. Sections 2 and 3 present materials and our methodology. The main results are showed and evaluated before concluding our paper.