Adaptive cuckoo search algorithm based fuzzy C means clustering with random walker algorithm for liver segmentation using CT images

• Published in: Multimedia tools and applications Vol. 84; no. 8; pp. 5051 - 5068
• Main Authors: Subha, S., Kumaran
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.03.2025; Springer Nature B.V
• Subjects: Adaptive algorithms; Algorithms; Clustering; Computed tomography; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Image segmentation; Liver; Medical imaging; Multimedia Information Systems; Search algorithms; Special Purpose and Application-Based Systems; OBL; Liversegmentation; Random Walker (RW); Fuzzy C-Means (FCM); Adaptive cuckoo search algorithm (ACSA)
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-024-18708-9

Liver segmentation from computed tomography (CT) images is a significant process for computer-aided diagnosis. Clustering is one of the efficient techniques for medical image segmentation. Although many clustering algorithms have been presented by the researchers, the Fuzzy C-Means (FCM) algorithm still provides better segmentation results. However, the performance of FCM clustering is further improved to avoid the resemblance of surrounding tissues'gray values in liver segmentation. Thus, an optimized FCM clustering with a Random Walker (RW) algorithm is proposed in this paper. Initially, the input liver CT images from the 3DIRCADB dataset are pre-processed. Then, the pre-processed images are given as input to the proposed clustering approach. In the proposed FCM clustering, cluster centers are chosen optimally using an adaptive cuckoo search algorithm (ACSA), in which the oppositional-based learning (OBL) technique is used to enhance the searchability of CSA. Besides, to manage the pixels or feature assignment in each cluster depending on the minima rule of segmentation, the RW algorithm is combined with the FCM. Simulation results depict that the proposed segmentation model attains a dice similarity coefficient (DSC) of 96.38% than the existing segmentation algorithms.