Performance analysis of EM-MPM and K-means clustering in 3D ultrasound image segmentation

Breast density is an important indicator for a woman´s lifetime risk of breast cancer. A 3D model of breast density can be obtained by taking 3D tomographic ultrasound and then identifying tissue distribution in the breast with 3D medical image segmentation. In this paper, we compare two segmentation algorithms, EM-MPM (Expectation Maximization with Maximization of Posterior Marginals) and K-means clustering using simulated phantoms. The computational phantoms cover various tissue density patterns. A total of twenty volumes of three dimensional synthetic ultrasound breast images were compared. We found that EM-MPM performs better than K-means Clustering on segmentation accuracy because the segmentation result fits the ground truth data very well. The EM-MPM is able to use a Bayesian prior assumption, which takes advantage of the 3D structure and finds a better localized segmentation. EM-MPM performs significantly better especially for the highly dense tissue scattered within low density tissue and for volumes with low contrast between high and low density tissues.