Automatic blood vessel-based liver segmentation through the portal phase abdominal CT dataset

摘要

Blood vessel has showed high performance at liver segmentation (rank 1 on the official website for liver segmentation). This method is semi-automatic including two stages: (1) semi-automatic abdominal blood vessels (ABVs) segmentation and subsequent classification into hepatic and non-hepatic, and (2) automatic liver segmentation based on segmented ABVs. In present paper, we propose to develop ABVs segmentation and classification to be fully automatic for full automation of the mentioned liver segmentation method. For automation of ABVs segmentation, average percentages of bone, kidneys, ABVs and abdominal part of the heart volumetries to the abdomen are estimated practically using MICCAI training data. These volumetries are used for automatic ABVs segmentation based on intensity, shape, volumetry and position information.