Comparative Analysis of K-Means Clustering and Watershed Segmentation Techniques for Brain Tumor Detection in MRI

This study aims to evaluate and compare the efficiency of two segmentation techniques, K-Means clustering, and Watershed segmentation, in the detection of brain tumors using MRI scans. Accurate segmentation of brain MRI images is crucial for the diagnosis and treatment of brain tumors. Numerous techniques have been employed for this purpose, but a clear comparison between these methods remains underexplored. This research provides a comparative analysis of K-Means clustering and Watershed segmentation to identify the superior technique for brain tumor detection. The study utilized MRI scans of patients with three different brain diseases: metastatic bronchogenic carcinoma, anaplastic astrocytoma, and sarcoma. Morphological operations were applied for skull removal, followed by segmentation into gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), and tumor areas using both K-Means clustering and Watershed segmentation. The segmented regions were quantified, and the efficiency of each technique was assessed through percentage calculations and visual representation via pie charts. The results indicate that K-Means clustering outperforms Watershed segmentation in accurately identifying and segmenting the tumor areas as well as the GM, WM, and CSF regions. For metastatic bronchogenic carcinoma, K-Means detected an average tumor area of 12.74%, compared to Watershed's 6.99%. Similar trends were observed for anaplastic astrocytoma and sarcoma, with K-Means consistently providing higher accuracy and clearer segmentation boundaries. K-Means clustering proves to be a more effective technique for brain tumor detection and segmentation in MRI scans compared to Watershed segmentation. The superior performance of K-Means is attributed to its ability to classify regions without the limitations associated with the structuring elements required by Watershed segmentation. Future research should further refine these techniques and explore their integration into automated diagnostic systems