Three-class Severity Detection and Classification of Brain Tumor ROI Using YOLO-v9

A vital part of diagnosing and treating medical issues is the use of medical imaging for the detection and categorization of brain tumors. For this research, we looked into how well the YOLOv9 deep learning architecture identified ROIs and classified brain tumors using a dataset consisting of nine thousand annotated images. One of the main aim of our work was to find the region of interest (ROI) in brain scans as precisely as possible so that cancers could be located. The second goal was to use the ROIs to categorize the tumors as meningioma, pituitary, or glioma. Mean average precision (mAP) for all tumor classes is 0.85, and for meningioma brain tumors particularly it is 0.949, indicating competitive accuracy metrics for the model. In this trained the network with weights from a generic object identification task before initializing it with YOLOv9 architecture-based transfer learning techniques. After that, adjusted the model based on the unique features of medical imaging by fine-tuning it using our dataset of brain tumors. The significant mAP scores, especially for meningioma tumors, highlight the potential of this deep learning approach to improve clinical workflows in radiology and neurology. This study contributes to the growing area of deep learning applications.