Inverse maximum likelihood-based edge detection for segmentation of breast lesion using active contour

Abstract

In this work, an attempt has been made to identify lesion in mammogram images using edge detection and segmentation method. The edge detection was carried out using inverse maximum likelihood ratio with pre-processing. The pre-processing includes removing of artefacts using top-hat transform and image enhancement using wavelet transform. The lesion edges are identified using inverse maximum likelihood ratio, and a binary mask is generated using active contour to segment the identified lesion. The generated mask is multiplied with the edge detected image to get the final segmented output. The performance of the segmentation method is analysed with ground truth and quantitative performance assessment is carried out using similarity measures based on region overlap and region statistics. These studies appear to be clinically relevant because automated analyses of lesion are important for breast cancer interventions.