YOLOV10 AND SAM 2.1 FOR ENHANCED MRI SEGMENTATION AND IMPROVED NEUROLOGICAL DISEASE DIAGNOSIS

Early and accurate diagnosis of neurological diseases through MRI imaging is crucial for effective treatment and patient management. This study presents a deep learning-based approach utilizing a diverse dataset of 12,121 MRI images spanning 12 categories across three major neurological diseases including Brain Tumor Disorders, Alzheimers Disease and Parkinsons Disease. The dataset was structured into 9,894 images for training and 2,227 for validation. Six YOLOv10 variants (N, S, M, B, L and X) were employed for multi-class classification and localization with the YOLOv10-X model achieving the highest diagnostic accuracy. To enhance interpretability the Segment Anything Model (SAM) 2.1 was applied for post-detection segmentation generating precise masks over detected regions further refined with plasma colormap visualization. Comparative evaluations highlight notable improvements in diagnostic performance demonstrating the effectiveness of integrating segmentation and explainable AI. This research contributes to the development of an advanced interpretable AI-driven framework for neurological disease detection.