Automatic Liver Segmentation Method based on Maximum A Posterior Probability Estimation and Level Set Method
Daisuke Furukawa, Akinobu Shimizu, Hidefumi Kobatake · 2007
Abstract. This paper describes an automatic liver segmentation method from three dimensional abdominal X-ray CT images. Our method ex-tracts liver regions based on the following two steps: (1) rough extraction based on maximum a posterior probability (MAP) estimation, and (2) precise segmentation using level set method. In the former process, the segmentation can be performed more accurately by using a combination of the probability density function approximated by the extended gaus-sian mixture distribution and a prior probability derived from a prob-abilistic atlas of liver. In the latter process, we introduce a novel term to prevent the level set method from extracting muscles as a part of the liver incorrectly. From the experimental results using ten test datasets distributed for the competition, it was confirmed that our method seg-mented the liver regions with volumetric overlap of about 88%. 1