Accurate and rapid brain tumor segmentation in Magnetic Resonance Imaging (MRI) remains a critical challenge for early diagnosis and effective treatment. Conventional methods often struggle with inherent tumor heterogeneity (variations in solid tumor, edema, necrosis) and noise, leading to segmentation inaccuracies. This work presents a novel hybrid approach specifically designed to address these limitations. It combines Entropy-driven Fuzzy C-Means (EnFCM) and Elephant Herding Optimization (EHO) algorithms. EnFCM, a novel contribution, utilizes information entropy to optimize the clustering of brain tissues and tumor sub-regions. This effectively handles the issue of overlapping intensities in MRI scans, where tissues might have similar intensity values. EHO, inspired by elephant herding behavior, further refines the cluster centers (centroids) identified by EnFCM. This iterative process enhances segmentation accuracy by ensuring optimal positioning within complex tumor regions. This combined EnFCM-EHO approach tackles limitations associated with noise and complex tumor characteristics, potentially leading to faster and more precise segmentation results. The proposed method's effectiveness is evaluated on the BraTS benchmark dataset.Keywords: Elephant Herding Optimisation (EHO), Entropy-based Fuzzy C- means clustering (EnFCM), Brain Tumour Segmentation (BraTS) dataset, Magnetic resonance imaging (MRI).