Development of semi-automatic liver segmentation method for three-dimensional computed tomography dataset

Loading...
Thumbnail Image
Date
2017-06
Authors
Chiang, Yi Fan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Segmentation of liver from 3D computed tomography (CT) dataset is very important in hepatic disease diagnosis and treatment planning. Manual segmentation gives accurate result but the process is tedious and time-consuming due to a large number of slices produced by the CT scanner. Low contrast of liver boundary with neighbouring organs, high shape variability of liver and presence of various liver pathologies will affect the accuracy of automatic liver segmentation and thus make automatic liver segmentation a challenging task. Therefore, a semi-automated liver segmentation program is developed in this project in order to obtain high accuracy in liver segmentation and reduce the time required for manual liver segmentation. The proposed algorithm can be divided into three stages. The first stage is parameter setup and pre-processing. User interaction is required to setup the segmentation parameters. For pre-processing, anisotropic diffusion filtering is applied to reduce noise in the image and smooth the image. In second stage, thresholding is applied to CT images to extract the possible liver regions. Then, morphological closing and opening are used close small holes inside liver region and break the thin connections between liver and neighbouring organs. Hole-filling is employed to fill up the large holes inside liver region. Next, the connected component analysis is performed to extract liver region from the CT slices. The last stage is post-processing. In post-processing, the contour of liver is smooth by binary Gaussian filter. The liver segmentation program with proposed algorithm is evaluated with CT datasets obtained from SLIVER07 to prove its effectiveness in liver segmentation. The results of liver segmentation achieved average VOE of 9.93 േ 4.36 %, average RVD of -0.03 േ 3.76 %, average ASD of 2.57 േ 1.73 mm, average RMSD of 5.82 േ 3.56 mm, and average MSD of 39.90 േ 17.23 mm. The total time required for the program developed to complete liver segmentation process is between 2 to 4 minutes. The proposed algorithm was able to segment the healthy liver effectively and efficiently even though there were over-segmentation and undersegmentation problem due to the presence of pathologies and low contrast between liver and neighbouring organs.
Description
Keywords
Citation