Kernel Based FCM for Spinal Cord Segmentation on Computed Tomography Images

Radiation therapy necessitates precise delineation of target structures such as the spinal cord, yet automated segmentation from CT images remains challenging due to complexities inherent in spinal cord anatomy and surrounding tissue. Addressing this, a novel approach leveraging Kernel-based Fuzzy C Means (KFCM) for spinal cord segmentation is introduced in this study. KFCM employs a kernel function to map data into a higher-dimensional space, facilitating effective separation within the feature space. This algorithm clusters voxels based on their distances from cluster centers in this transformed space, thereby enabling accurate segmentation. The proposed method is rigorously evaluated using a dataset comprising 20 CT images, with performance benchmarked against two leading segmentation algorithms. Notably, the results demonstrate superior segmentation accuracy and computational efficiency of the KFCM algorithm. Through qualitative analysis, it is evident that the KFCM-based segmentation yields finer delineation of the spinal cord, even amidst complexities introduced by materials like the spinal canal. Quantitative assessment further underscores the efficacy of the proposed approach, with commendable segmentation accuracy values observed. These findings underscore the potential of the KFCM algorithm as a robust tool for medical image segmentation, particularly in applications demanding precise delineation of anatomical structures for radiation therapy planning. The successful deployment of this methodology signifies a significant advancement in medical image analysis, offering clinicians a reliable tool for enhancing treatment planning and patient outcomes. Moving forward continued refinement and validation of the KFCM algorithm hold promise for broader integration into clinical practice, ultimately contributing to improved healthcare delivery and outcomes for patients undergoing radiation therapy.