Abstract
Alzheimer's disease (AD) is a progressive neurodegenerative disorder and the leading cause of dementia worldwide, affecting over 55 million people globally, with numbers expected to rise dramatically. Early detection and classification of AD are crucial for improving patient outcomes and slowing disease progression. However, conventional diagnostic approaches often fail to provide accurate classification in the early stages. This paper proposes a novel approach using advanced computer-aided diagnostic (CAD) systems and the YOLOv11 neural network for early detection and classification of AD. The YOLOv11 model leverages its advanced object detection capabilities to simultaneously localize and classify AD-related biomarkers by integrating multimodal data fusion of T2-weighted MRI and DTI images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Regions of interest (ROIs) were selected and annotated based on known AD biomarkers, and the YOLOv11 model was trained to classify AD into four stages: Cognitively Normal (CN), Early Mild Cognitive Impairment (EMCI), Late Mild Cognitive Impairment (LMCI), and Mild Cognitive Impairment (MCI). The model achieved exceptional performance, with 93.6% precision, 91.6% recall, and 96.7% mAP50, demonstrating its ability to identify subtle biomarkers by combining MRI and DTI modalities. This work highlights the novelty of using YOLOv11 for simultaneous detection and classification, offering a promising strategy for early-stage AD diagnosis and classification.