A comparative study of ultrasound image segmentation algorithms for segmenting kidney tumors

Abstract

In this paper we introduce an ultrasound image segmentation evaluation framework for kidney tumor. Ultrasound image segmentation algorithms can be divided into edge based, region based, texture based, active contour and model base technique. We tested the performance of algorithms in each category using a kidney phantom and kidney cyst ultrasound image. We found that the algorithms we implemented are more suitable for relatively homogeneous kidney tumors. For more heterogeneous tumors we should use more complicated segmentation techniques and some of these advanced techniques are discussed in this paper.