Intelligent Multi-Class Brain Tumor Detection and Localization Using Majority Voting and Object Detection

Deep learning approaches have emerged as powerful tools for automating brain analysis for identification and precise localization of brain tumors, which is important in treatment planning. In this study, a robust ensemble-based framework is proposed for brain tumor detection and localization in magnetic resonance imaging (MRI) scans. A majority voting strategy is employed to fuse the predictions of three complementary deep learning models: a multi-channel convolutional neural network (CNN), a hybrid CNN combined with a support vector machine (SVM) classifier, and a YOLO-based object detection network. Experimental results demonstrate that the ensemble strategy significantly improves tumor detection and classification performance compared to individual models. The proposed method achieves an accuracy of ۹۹% and a precision of ۱۰۰%, indicating its high robustness and consistency. Furthermore, the integration of the YOLO-based detector enables accurate spatial localization of tumor boundaries. The object detection component specifically achieves an F۱-score of ۹۲.۲۵%, and a mean average precision (mAP) of ۹۴.۹% at IoU threshold ۰.۵ (mAP۵۰) and ۶۹.۷% across IoU thresholds ۰.۵–۰.۹۵ (mAP۵۰–۹۵). These results confirm the model’s effectiveness in delineating tumor regions with high spatial fidelity

Keywords:

Brain Tumor , Convolutional Neural Network (CNN) , Support Vector Machine (SVM) Classifier , Majority Voting , YOLO Object Detection Model