Self-organizing maps for masking mammography images

This paper describes a new image segmentation algorithm for masking the breast region from the background in digital mammograms. The algorithm is applied to 160 images and shows promising results. Evaluation is based on comparisons with a histogram/region-growing algorithm. A self-organizing map is used to obtain an initial segmentation. The weight vectors of the self-organizing map are then clustered using the K-means method. Knowledge-based refinement provides the final binary mask that segments the image. Results indicated that the proposed approach could be used as the first stage in a computer-aided diagnostic system.