An Application for Computer-Assisted Automatic Segmentation of Liver on Computed Tomography Images

Abstract

In recent years, thanks to the development of imaging techniques, computer aided detection (CAD) systems have become widely used in medical image segmentation. The most important step in CAD image processing applications is to perform segmentation with high accuracy. In this study, an automated computer-assisted method and an application software for the segmentation of the liver on abdominal computed tomography (CT) images are developed. For the segmentation of the liver region, region growing (RG) and fuzzy C-means (FCM) algorithms are used. In order to measure the performance of segmentation with these algorithms, a physician marked out the boundary of the liver. In the study, the segmentation results obtained by the RG and FCM algorithms are compared on the developed application software using physician selection criteria. Jaccard similarity criterion are used to compare segmentation results. In the experimental studies on 88 CT images, average performance values are obtained for 91.15% in the RG algorithm and 75.16% in the FCM algorithm according to the Jaccard similarity criterion. As a result, segmentation with the RG algorithm is more successful. In addition, the statistical significance of quantitative values obtained from similarity measurements is measured. It is concluded that the segmentation results obtained by the RG algorithm revealed a significant difference by evaluating the significance tests. Moreover, the segmentation process times are compared with both segmentation methods and segmentation with RG is found to be faster. The findings show that the proposed method can be used as a secondary tool in the decision-making process of physicians.

Keywords

• Liver
• CT Imaging
• Computer-Assisted Detection
• Image Segmentation
• Fuzzy C-Means
• Region Growing