Advancing Brain Tumor Detection with YOLOv9: A Comprehensive Evaluation

• Published in: 2025 4th International Conference on Computing and Information Technology (ICCIT) pp. 439 - 444
• Main Authors: Elnady, Norhan, Adel, Aya, Badawy, Wael
• Format: Conference Proceeding
• Language: English
• Published: IEEE 13.04.2025
• Subjects: Accuracy; automated diagnostic systems; Biomedical imaging; Brain modeling; Brain Trumer detection; Brain tumors; image augmentation; Machine learning algorithms; Magnetic resonance imaging; Network architecture; object detection in medical imaging; Planning; Training; YOLO; YOLOv9
• DOI: 10.1109/ICCIT63348.2025.10989289

This paper presents a comprehensive study on the application of the latest YOLO (You Only Look Once) version, YOLOv9, for detecting brain tumors using magnetic resonance imaging (MRI) data. Building on the successes of previous YOLO models such as YOLOv3, YOLOv5, and YOLOv8, which have been effectively used in brain tumor detection tasks, our study explores the enhancements introduced in YOLOv9. These enhancements aim to improve the accuracy, efficiency, and speed of tumor localization and detection. YOLOv9 introduces advanced architectural modifications, including an updated backbone, enhanced feature pyramids, and a sophisticated training algorithm that leverages state-of-the-art machine learning techniques. We conducted a series of experiments to evaluate the model's performance, focusing on detection accuracy, mean average precision (mAP), and recall values, comparing them with those of its predecessors. The results indicate that YOLOv9 achieves a higher mAP of 98.2%, demonstrating significant improvements over YOLOv8. This paper provides a detailed analysis of YOLOv9's architecture, its implementation for brain tumor detection, and a discussion on its potential implications for future research and clinical applications.