AI-BASED EARLY DETECTION OF ALZHEIMER'S DISEASE USING MULTIMODAL MRI: A SYSTEMATIC REVIEW OF DEEP LEARNING APPROACHES

Abstract

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder globally, affecting approximately 55 million individuals and projected to triple by 2050. Early and accurate diagnosis remains a critical clinical challenge, as irreversible neurodegeneration precedes symptom onset by 15–20 years. Deep learning applied to multimodal MRI — integrating structural, functional, and diffusion tensor imaging — has demonstrated transformative diagnostic potential, achieving diagnostic accuracy of 85–95% in distinguishing AD from mild cognitive impairment and healthy controls in controlled research settings. This review systematically analyses 47 peer-reviewed studies published between 2018 and 2024, evaluating convolutional neural network architectures, transformer-based models, and graph neural networks for AD detection from MRI data. We identify optimal model architectures, training dataset requirements, and critical implementation challenges including data harmonisation across MRI scanners, class imbalance, and clinical validation gaps. The findings provide a roadmap for translating deep learning MRI diagnostics from research to clinical practice in neurology departments with limited specialist capacity.