VSFAS

vsfas − segments fat, skin, and reconstruction boundary of the body

fatandskinsegmentation if=<infile> rt=<root> of=<> os=<> oi=<> oa=<> ond=<>

vsfas segments the fat and skin from a CT image. The input must be a CT image since a large portion of the processing relies on thresholds related to tissue values. Each output is in the form of a binary image.

The skin is found by finding the entire body, then tracing the boundary of the body that touches air outside of the body. The reconstruction boundary is found by tracing the portions of the body that touch the reconstruction limits of the scanner. The points in the image outside of the reconstruction boundary are typically set to a value less than -1000 HU. The skin and the reconstruction boundary output form a surface that completely encloses the body.

The fat segmentation is sensative to noise levels in the image. Different levels of filtering are used for the fat segmentation based on a local estimate of noise in a 15x15x1 area of a homogenous region of tissue. The outcome of any filtering is determined to be fat if it is between -190 and -30 HU and not in a high gradient area. The exclusion of high gradient area prevents vessels in the lung from being segmented as fat since they are mostly partial voxels. The air and no-data regions are also extracted. The air is defined as air outside of the body. Hence this region will not include lung or airways. The no-data region is defined as the region in which the scan is not reconstructed. In most cases this is the rectangular prism in which the cylinder of data is inscribed.

Input file must be a CT image.

J. Padgett