YOLO-MEDX: efficient brain tumor detection with YOLO11-EfficientNet

We introduce YOLO-MEDX, a refined object detection model built upon the YOLO11 framework and EfficientNet, aiming to enhance the accuracy and robustness of intracranial lesion localization in medical imaging. The model effectively tackles key challenges like small lesion size, low image contrast, and complex visual environments by incorporating attention modules, lightweight convolution layers, and regularization techniques into its backbone. Experiments on a brain tumor dataset show that YOLO-MEDX achieves 94.41% mAP@0.5 and 69.57% mAP@0.5:0.95, outperforming the baseline YOLO11 model. Detection precision for Meningioma, Glioma, and Pituitary Tumors reaches 97%, 85%, and 95%, respectively. The normalized confusion matrix further validates the model's ability to differentiate between lesion categories while minimizing background noise. This demonstrates that carefully restructured architectures can significantly boost detection performance in medical images without heavy reliance on intricate module-level optimization. YOLOMEDX thus provides a compelling strategy for reliable and efficient AI-driven diagnostics.