Glioma Detection Using YOLO V9: A Deep Learning Framework

Glioma, a type of brain tumor, poses a serious challenge to patient survival, calling for early and fast detection. Manual detection is time-consuming and prone to mistakes, necessitating effective automated solutions. Recent advancements in deep learning enabled promising computer-aided diagnosis (CAD) systems using Magnetic Resonance Imaging (MRI), offer for detecting brain tumors. This study proposes a novel deep transfer learning approach for glioma detection using the YOLOv9 object detection framework, which is leading solution for real-time object detection by integrating the Generalized Efficient Layer Aggregation Network (GELAN) in its architecture to offer a balance between speed, accuracy, and adaptability. This study utilizes an MRI dataset that was released in Nov 2024. Our proposed approach based on the YOLOV9 tiny model produces a precision of 0.816, recall of 0.767, mAp @ 50 of 0.813 and mAp50-95 of 0.41 for glioma detection using 2-classes of objects i.e. “Glioma” and “No-Glioma”. These results indicate that YOLOV9 can be an effective CAD system. Future work will focus on choosing a local dataset for further optimizing the model.