Modified Density-Based Data Clustering for Interactive Liver Segmentation

Identifying the liver region from abdominal Computed Tomography (CT) scans is still a challenging task due to the complexity of the liver's anatomy, similar intensity with adjacent organs and presence of pathologies. In this paper we propose a system which consists of three stages: preprocessing, segmentation and post processing. In the first stage, the input image is smoothed using anisotropic diffusion filter. In the second stage, we introduce a modified density-based clustering algorithm, DBSCAN, to segment the liver. Finally, morphological operations are performed to enhance the segmentation results. The proposed system is evaluated on 3Dircadb1 database which is publicly available. The experimental results show that the proposed system is effective for accurate detection of the liver surface in comparison with other related works in the literature.