Influence of distance transform on classification based segmentation of abdominal organs from MRI

Segmentation of abdominal organs from MR images can be performed using two- or three- dimensional (2D or 3D) methods. 2D methods are less sensitive to volumetric anisotropy than 3D methods. Also they have the advantage of direct integration to the clinical scheme based on manual delination. One of the most important factors that affects the performance of 2D methods is the extraction and usage of similarities between adjacent image slices. In this paper, the effects of the use of distance transform as adjacent slice similarity indicator on performance is discussed. Considering classification based segmentation of abdominal organs, images modified by distance transform are used to extract features. Application of the same classification strategy to modified and original images show the advantage of the usage of the distance transform on performance.