A segmentation algorithm of 3D ultrasonic data based on tissue characterization

In a previous work [1], a segmentation algorithm was proposed for 3D ultrasonic data. The algorithm involves 3D adaptive K-Means clustering of the gray-scale and texture features images calculated from the envelope image. The segmentation problem was formulated as a Maximum A Posterior (MAP) estimation problem. The method was demonstrated successfully on in vivo breast data with texture features calculated on the cooccurrence matrices. A major difficulty in the proposed algorithm is the choice of the texture features which characterize the different tissues. In the case of ultrasonic data, two major classes of parameters exist, acoustical and textural parameters. In this work both acoustic and texture characterization are taken into account in the segmentation process.