AUTOMATIC LIVER SEGMENTATION FROM CT SCANS USING INTENSITY ANALYSIS AND LEVEL-SET ACTIVE CONTOURS

Abstract

Liver segmentation from CT scans is still a challenging task due to the liver characteristics in terms of shape and intensity variability. In this work, we propose an automatic segmentation method of the liver from CT data sets. The framework consists of three main steps: liver shape model localization, liver intensity range estimation and localized active contouring. We proposed an adaptive multiple thresholding technique to estimate the range of the liver intensities. First, multiple thresholding is used to extract the dense tissue from the whole CT scan. A localization step is then used to find the approximate location of the liver in the CT scan, to localize a constructed mean liver shape model. A liver intensity-range estimation step is then applied within the localized shape model ROI. The localized shape model and the estimated liver intensity range are used to build the initial mask. A level set based active contour algorithm is used to deform the initial mask to the liver boundaries in the CT scan. The proposed method was evaluated on two public data sets: SLIVER07 and 3D-IRCAD. The experiments showed that the proposed method is able to segment to liver in all CT scans in the two data sets accurately.

Keywords

• Automatic segmentation
• Intensity analysis
• Localized contouring
• Multiple thresholding