Multi-Class Classification and Detection of Brain Tumor ROI: Yolov9-CNN

The detection and classification of brain tumors through medical imaging play a critical role in the identification and management of medical conditions. This study focused on two primary objectives: first, to accurately identify the ROI within brain scans, enabling precise localization of tumors, and second, to classify the tumors into specific types—meningioma, pituitary, and glioma—based on the identified ROIs. To achieve this, this study proposes a YOLOv9 deep learning architecture with a transfer learning hybrid framework. Proposed method began by initializing the network with pre-trained weights from a general object detection task. This model achieved competitive accuracy metrics, with an average mean average precision (mAP) of 0.85 for all tumor classes and 0.949 specifically for meningioma brain tumors. With six different Yolo models, the proposed methodology shows enhanced performance in the metrics. This research adds to the expanding field of deep learning applications in medical image analysis, showcasing opportunities for enhancing brain tumor detection and classification through advanced computer vision techniques.