Abstract
Alzheimer’s disease is a progressive neurodegenerative condition that leads to cognitive decline, memory loss, and ultimately loss of independence. Early detection of Alzheimer’s disease is essential for timely intervention, but traditional diagnostic tools often fall short due to their reliance on manual evaluation and limited sensitivity. This study introduces a novel application of the latest advancement in real-time object detection framework to detect and localize biomarkers of Alzheimer’s disease using structural brain magnetic resonance imaging data. The raw dataset—sourced from publicly available repositories such as Alzheimer’s Disease Neuroimaging Initiative contains labeled MRI scans categorized into Alzheimer’s Disease, Mild Cognitive Impairment, and Cognitively Normal classes. Extensive preprocessing steps were conducted, including image normalization, resizing, and the removal of corrupted scans. To mitigate class imbalance, data augmentation techniques such as brightness modulation, horizontal flipping, and rotation were selectively applied to underrepresented classes like Mild Cognitive Impairment. Further, bounding box annotations were used to highlight critical brain regions, enabling cutting-edge computer vision model developed by Ultralytics to localize and classify abnormal patterns effectively. The model was trained with transfer learning and fine-tuning strategies to enhance diagnostic precision while maintaining computational efficiency. Quantitative evaluation through precision-recall curves, F1-confidence analysis, and confusion matrices demonstrates that real-time object detection technology is highly capable of distinguishing among Alzheimer’s Disease, Mild Cognitive Impairment, and Cognitively Normal cases. The model achieved a mean Average Precision of 0.552, underscoring its robustness. The integration of localization and classification within a real-time, interpretable framework presents real-time object detection as a promising tool for scalable and non-invasive Alzheimer’s disease screening.