A Novel Post-Process Approach for Fast Marching Method in Liver CT Slices Automatic Segmentation

Liver segmentation on computed tomography (CT) slices is a challenging task because the images are often corrupted by noise and sampling artifacts. Recent years fast marching method (FMM) has been introduced into the image segmentation domain and proved to have advantage in blur edge detection. When apply the FMM to the segmentation of liver CT slice, to attain the completely liver shape, an over-segmentation result was usually unavoidable due to the poor contrast between the liver matter and the surrounding tissues. Therefore, in this paper, based on the FMM, a novel post-process approach is proposed. This approach mainly depends on the curve-fitting algorithm and takes into account the liver shape continuity and comparability. Followed with this post-processing, our algorithm can segment the liver CT slices correctly and quickly. First, a speed image can be generated after pretreatment such as filtering and noise reduction. Second, according to the characteristics of liver CT slices, the FMM parameters are attained from contiguous slice to continue the segmentation procedure. Finally, liver boundary is corrected by our approach. The whole procedure is nearly complete automatization, only a seed point is needed at the beginning.