Automated Detection and Spatial Localization of Alzheimer’s Disease Stages from MRI Using Deep Learning–based Object Detection

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline, memory impairment, and distinct structural brain alterations observable in magnetic resonance imaging (MRI). Early and accurate diagnosis remains a critical challenge, as conventional clinical assessments and manual image interpretation are time-consuming and subject to inter-observer variability. This study presents an automated deep learning framework for the detection and spatial localization of Alzheimer’s disease stages from MRI using the YOLOv11 family of object detection models. Two variants, YOLOv11N and YOLOv11S, were implemented and evaluated using two-dimensional slices extracted from three-dimensional T1-weighted MRI scans obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. The models classify MRI slices into three diagnostic categories: Alzheimer’s Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN). Experimental results demonstrate that YOLOv11N, the lightweight “Nano” variant, consistently outperformed YOLOv11S and is therefore identified as the best-performing and recommended model, while YOLOv11S serves primarily as a baseline and comparative reference. Trained with extensive data augmentation to address class imbalance, YOLOv11N achieved a precision of 0.7315, recall of 0.7378, and a mean Average Precision (mAP@0.5) of 0.8145, exceeding the performance of the larger YOLOv11S model while requiring substantially fewer computational resources. In addition to multi-class classification, the proposed framework provides spatial localization of AD-related structural abnormalities, enhancing clinical interpretability by visually highlighting affected brain regions within MRI slices. These results indicate that YOLOv11N offers an effective balance of accuracy, interpretability, and efficiency, making it suitable for clinical use to support earlier diagnosis, monitor disease progression, and guide personalized patient care.