Advanced Brain Tumor Classification Utilizing MaskR-CNN with Fuzzy C-Means Clustering

Abstract Brain tumors are a major concern in medical diagnostics, necessitating precise and prompt detectionfor optimal treatment. While radiologists may diagnose them using medical imaging, automating theprocess has several advantages, including increased efficiency and accessibility. Magnetic resonanceimaging (MRI) is the most effective way to detect these tumors. An automated system that accuratelyidentifies and localizes them in MRI scans can greatly benefit patients lacking immediate access toa radiologist. Our study proposes a method combining Mask R-CNN and Fuzzy C-Means (FCM)clustering to boost the accuracy of brain tumor detection in MRI images. This approach identifiestumors and segments abnormal brain tissues but also gauges the likelihood of tumor presence, offering acomprehensive diagnostic tool. While Mask R-CNN excels in segmentation accuracy, integrating FCMclustering further refines the detection process. FCM, which allows data to belong to multiple clusters,is proven in medical images where tumor boundaries can be ambiguous. Our experiment utilizes theBraTS 2018 and 2020 datasets, employing five-fold cross-validation to demonstrate the effectiveness andvalidate the clinical application of our proposed method.