YOLOv11-Based Brain Tumor Detection and Segmentation for Clinical Decision Support

Brain tumors pose a severe threat to human lives, with cases rising globally. Every year, approximately 250,000 individuals worldwide are diagnosed with brain tumors, contributing to a mortality rate exceeding 150,000 deaths annually. Late diagnosis significantly reduces survival rates, whereas early detection and timely intervention can improve patient outcomes by up to 75 %. To address this critical challenge, researchers are increasingly employing machine vision-based deep learning techniques for brain tumor detection and segmentation from MRI scans. In this research, we present an enhanced deep learning-based system utilizing YOLOv11 for real-time brain tumor detection and segmentation. The model incorporates a CSPDarkNet backbone, Spatial Pyramid Pooling Fast (SPPF) neck, and an anchor-free detection head, optimizing both detection accuracy and computational efficiency. Our system processes MRI scans in real-time, accurately localizing and segmenting tumors, measuring their size in cm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, and identifying multiple tumors with cumulative area computation. The model was trained on 834 annotated MRI images, augmented to 1960 samples to improve generalization. Experimental evaluation demonstrates the system’s effectiveness, achieving 97.6% precision, 99.2% recall, 99.46% mAP@50, and 99.46% segmentation IoU, outperforming previous YOLO-based approaches in both accuracy and processing speed. The model is seamlessly integrated into a PyQt5-based graphical user interface (GUI), enabling non-expert clinical operators to interact effortlessly with the system. Due to its high precision, real-time processing capability, and user-friendly design, this system is a viable solution for assisting radiologists in brain tumor diagnosis and treatment planning, potentially reducing mortality rates and enhancing patient survival.