An Optimized Hybrid Active Contour Model for Liver CT Image Segmentation

Liver segmentation from abdominal Computed Tomography (CT) scans is a critical preliminary step in computer-aided diagnosis, tumor detection, and surgical planning. However, this task remains highly challenging due to fuzzy organ boundaries, low contrast, and interference from adjacent organs with similar intensities. This study aims to propose an automated and robust hybrid framework to overcome these limitations. The proposed methodology integrates local clustering with swarm intelligence to optimize an energy-based segmentation model. Initially, after noise reduction and contrast enhancement, a local K-Means thresholding technique is applied to extract a reliable preliminary liver mask. Subsequently, the Social Spider Optimization (SSO) algorithm is employed to dynamically search the feature space-incorporating texture, edge, and intensity parameters-to optimize the Active Contour model. This swarm-based global search prevents the active contour from getting trapped in local minima or leaking into non-liver tissues. The proposed approach was evaluated on the standard SLIVER۰۷ dataset. Quantitative results demonstrate that the hybrid method achieves an outstanding Accuracy of ۹۸.۹۷%, a Sensitivity of ۹۱.۲۵%, and a Specificity of ۹۸.۵۰%. The findings prove that the SSO-driven Active Contour model significantly outperforms recent state-of-the-art segmentation techniques, particularly in enhancing sensitivity and accurately preserving complex liver morphologies without over-segmentation.

کلیدواژه ها:

Liver Segmentation ، Active Contour ، Social Spider Optimization ، K-Means Clustering