Introduction - If you have any usage issues, please Google them yourself
This paper proposes and evaluates a promising technique
used for liver segmentation that can be used as the first step in
liver treatment planning. It is based on texture feature
analysis obtained from 3D DCT block coefficients, support
vector machines, and refinement provided by combined
morphological operations, median filtering and connected
components. The entire segmentation process was conducted
at a 3D level (both classification and post processing step)
resulting in a faster computation time than existing 2D
approaches. The obtained results on real CT volumes have
demonstrated efficiency of the proposed approach.
Furthermore, if the 3D-DCT based volume compression will
be standardized, the proposed features will be available
directly, at no computational cost. Further work will try to
improve the segmentation algorithm, including also some
spatial information or liver masks.