Automatic Liver Segmentation from CT Scans Using Multi-layer Segmentation and Principal Component Analysis

This paper describes an automatic liver segmentation algorithm for extracting liver masks from CT scan volumes. The proposed method consists of two stages. In the first stage, a multi-layer segmentation scheme is utilized to generate 3D liver mask candidate hypotheses. In the second stage, a 3D liver model, based on the Principal Component Analysis, is created to verify and select the candidate hypothesis that best conforms to the overall 3D liver shape model. The proposed algorithm is tested for MICCAI 2007 grand challenge workshop dataset. The proposed method of this paper at this time stands among the top four proposed automatic methods that have been tested on this dataset.