A Parallelized Grey Wolf Optimizer-Based Fuzzy C-Means for Fast and Accurate MRI Segmentation on GPU

Magnetic Resonance Imaging (MRI) has a pivotal role in medical image analysis, for its ability in supporting disease detection and diagnosis. Fuzzy C-Means (FCM) clustering is widely used for MRI segmentation due to its ability to handle image uncertainty. However, the latter still has countless limitations, including sensitivity to initialization, susceptibility to local optima, and high computational cost. To address these limitations, this study integrates Grey Wolf Optimization (GWO) with FCM to enhance cluster center selection, improving segmentation accuracy and robustness. Moreover, to further refine optimization, Fuzzy Entropy Clustering was utilized for its distinctive features from other traditional objective functions. Fuzzy entropy effectively quantifies uncertainty, leading to more well-defined clusters, improved noise robustness, and better preservation of anatomical structures in MRI images. Despite these advantages, the iterative nature of GWO and FCM introduces significant computational overhead, which restricts their applicability to high-resolution medical images. To overcome this bottleneck, we propose a Parallelized-GWO-based FCM (P-GWO-FCM) approach using GPU acceleration, where both GWO optimization and FCM updates (centroid computation and membership matrix updates) are parallelized. By concurrently executing these processes, our approach efficiently distributes the computational workload, significantly reducing execution time while maintaining high segmentation accuracy. The proposed parallel method, P-GWO-FCM, was evaluated on both simulated and clinical brain MR images, focusing on segmenting white matter, gray matter, and cerebrospinal fluid regions. The results indicate significant improvements in segmentation accuracy, achieving a Jaccard Similarity (JS) of 0.92, a Partition Coefficient Index (PCI) of 0.91, a Partition Entropy Index (PEI) of 0.25, and a Davies-Bouldin Index (DBI) of 0.30. Experimental comparisons demonstrate that P-GWO-FCM outperforms existing methods in both segmentation accuracy and computational efficiency, making it a promising solution for real-time medical image segmentation.