A modified graph cuts image segmentation algorithm with adaptive shape constraints and its application to computed tomography images. (September 2020)

• Record Type: Journal Article
• Authors: Chen, Hong; Pan, Xianpan; Lu, Xuesong; Xie, Qinlan
• Is Part Of: Biomedical signal processing and control. Volume 62(2020)
• Publication Date: 2020-09
• Subjects: Graph cuts -- Image registration -- Shape prior -- Computed tomography image; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.102092
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A multi-object segmentation method was proposed to extract the specific image regions pertaining to individual target organs in abdominal CT images. The proposed method is focus on address these problems of low contrast, blurred edges, and irregular contours of organs such as the liver, kidney, and spleen in abdominal CT images. The proposed method combines the modified graph cuts algorithm with the adaptive shape constraints, which is implemented by adding a constraint energy term of corresponding shape prior to the graph cuts energy function. With the method, the originaly segmentation by multi-atlas registration is employed as a shape prior to constrain segmentation result. Then the images of individual target organs are extracted by minimizing the energy function using the maximum-flow minimum-cut algorithm. The levels of shape constraint applied to the graph cuts energy function are adjusted according to differences in the probability of adjacent pixels residing within the target region, which are obtained in the initial segmentation process. Experimental results demonstrate that the proposed method can segment target organs well and can effectively reduce the occurrences of over-segmentation and under-segmentation caused by the conventional graph cuts algorithm. Abstract: The low contrast, blurred edges, and irregular contours of organs, such as the liver, kidney, and spleen, in abdominal computed tomography (CT) images hamper the machine-aided extraction of specific image regions pertaining to individual target organs. This problem is addressed herein by proposing an improved graph cuts segmentation algorithm with adaptive shape constraints. First, the original image is segmented based on multi-atlas registration. Then, the segmentation result is employed as a shape prior to constrain the shape of the final graph cuts segmentation result by adding a shape constraint energy term to the graph cuts energy function. The levels of shape constraint applied to the graph cuts energy function are adjusted according to differences in the probability of adjacent pixels residing within the target region, which are obtained in the initial segmentation process. Finally, the images of individual target organs in abdominal CT images are extracted by minimizing the energy function using the maximum-flow minimum-cut algorithm. Experimental results demonstrate that the proposed method can segment target organs well and can effectively reduce the occurrences of over-segmentation and under-segmentation caused by the conventional graph cuts algorithm.