Segmentation of Liver from CT Abdominal Images

Abstract

Automatic segmentation of liver from the abdominal Computed Tomography images is a difficult task. It is very important to segment the liver accurately, so the tumors can be located, detected and classified accurately within a liver. The proposed segmentation methods include preprocessing stage as first step where image resizing and grayscale conversion is performed. Thresholding technique is applied to obtain a binary image. Next, liver is segmented from 2-D abdominal CT scanned images using various segmentation methods like adaptive thresholding with morphological operations, global thresholding with morphological operations and Watershed gradient transform. Next, Active contour balloon snake model is applied on 3-D dataset 3D-IRCADb (3D Image Reconstruction for Comparison of Algorithm Database). The empirical comparative study is carried out using JSC, DSC, sensitivity, specificity and accuracy and results are tabulated. The empirical comparative study of these methods using Dice and Jaccard Similarity Coefficient is carried out and results are tabulated.