A LIGHTWEIGHT ATTENTION-ENHANCED YOLOV8M FRAMEWORK FOR BRAIN TUMOR DETECTION USING MRI IMAGES

Abstract

Brain tumors are one of the most serious neurological diseases and early detection is essential for effective treatment and prognosis. Magnetic Resonance Imaging (MRI) is commonly used to diagnose brain tumors; nevertheless, manual interpretation is time-consuming and requires extensive radiological skill. In recent years, significant advances in deep learning have been achieved, which have led to the improvement of the accuracy and efficiency of automated brain tumor diagnosis. This work proposes an upgraded lightweight attention-guided YOLOv8m model with an Enhanced Spatial Attention (CBAM) module to enhance feature extraction and tumor localization, while maintaining computational efficiency. The proposed model was trained and evaluated with the publicly available Figshare brain MRI dataset containing glioma, meningioma and pituitary tumor images. To enhance the model's resilience.generalization, standard preprocessing, data augmentation, and transfer learning methods were applied. Experimental results demonstrate that the proposed method achieves better performance than the baseline YOLOv8m model and is comparable to the existing attention-based YOLOV8m framework. The attention mechanism boosts the detection rate of tumors and reduces false positives, making the model suitable for real-time computer-aided diagnosis. The proposed framework is an efficient and reliable solution to automate brain tumor diagnosis and can be used to support radiologists in clinical decision making.