A Hybrid Approach for Brain Tumor Segmentation Using Fuzzy C-Means and the Imperialist Competitive Algorithm
| Author | Jafar Emamipour | en |
| Author | Hamzeh Vahidifar | en |
| Author | Hossein Nahid-Titkanlue | en |
| Accessioned Date | 2026-07-14T18:05:58Z | |
| Issued Date | 2026-03-31 | en |
| Abstract | Background: Accurate, automated segmentation of brain tumor regions from magnetic resonance imaging (MRI) is critical for computer-aided diagnosis and radiotherapy planning. Among unsupervised techniques, Fuzzy C-Means (FCM) is widely used because it can accommodate data uncertainty and partial volume effects. However, standard FCM is highly sensitive to the random initialization of cluster centers, which can lead to convergence to local optima and reduced segmentation reproducibility. Objectives: This study aimed to develop a robust hybrid segmentation framework to mitigate the initialization sensitivity of conventional clustering. The primary objective was to leverage the global optimization capability of the Imperialist Competitive Algorithm (ICA) to identify optimal initial cluster centers, thereby ensuring stable convergence and improving segmentation accuracy. Methods: The proposed method integrates ICA with FCM, thereby replacing the conventional random initialization strategy. ICA mimics sociopolitical competition to explore the search space globally and identify near-optimal cluster centroids, which are then used as initial inputs for the subsequent local FCM refinement phase. The framework was evaluated using the BraTS 2024 dataset. Performance was validated against K-means, standard FCM, and Kernel-based Intuitive Fuzzy C-Means (KIFCM) using accuracy, sensitivity, and precision. Results: The experimental analysis showed that the proposed FCM-ICA framework significantly outperformed the comparative algorithms. With an optimal population size of 50 countries, the method achieved a classification accuracy of 88.1% and a sensitivity of 85.5%. In comparison, K-means, standard FCM, and KIFCM yielded accuracies of 66.3%, 82.0%, and 86.9%, respectively. ICA-based initialization effectively mitigated the local optimum problem and provided greater stability than random initialization. Conclusions: Integrating ICA to optimize cluster centers significantly improves fuzzy clustering performance in medical imaging. The proposed hybrid method provides a robust, accurate, and stable solution for brain tumor segmentation and demonstrates strong potential for integration into automated clinical diagnostic workflows. | en |
| DOI | https://doi.org/10.69107/tatm-170560 | en |
| URI | https://brieflands.com/journals/tatm/articles/170560 | en |
| URI | https://repository.brieflands.com/handle/123456789/68002 | |
| Keyword | Image Segmentation | en |
| Keyword | Brain Tumor MRI | en |
| Keyword | Fuzzy Clustering Algorithm | en |
| Keyword | Imperialist Competitive Algorithm | en |
| Publisher | Brieflands | en |
| Title | A Hybrid Approach for Brain Tumor Segmentation Using Fuzzy C-Means and the Imperialist Competitive Algorithm | en |
| Type | Research Article | en |