Liver Segmentation Based on Hybrid-Distance Regularized Level Set Evolution Combining Region Growing in Abdominal Computed Tomography

Liver segmentation from computed tomography (CT) images is a prerequisite for various computer-aided clinical applications. However, it is still a challenging task due to the low contrast of adjacent organs and the varying shapes between subjects. This paper presents an interactive liver segmentation method based on a new hybrid level set and region growing. Firstly, in the image preprocessing step (image denoising, specific proportional gradient filter to enhance the liver edge, nonlinear gray-scale conversion and custom Binary conversion), the original abdominal CT are converted to binary images. Next, a small number of seed points are set for the region growing to roughly extract the liver region. Then, the segmentation result is optimized by the hybrid level set, which is driven by the image edge information and the region information, thus it can adapt to the larger balloon force with more iterations and better ability to resist edge leakage, compared with single image-driven level set. The method was validated on 40 sets of datasets provided from cooperative hospital and 20 sets from public datasets 3D-IRCDb-01, and the results showed that compared with other methods, our proposed method can achieve higher segmentation accuracy.