Multimodal data fusion with artificial intelligence for Alzheimer’s disease diagnosis: A comprehensive review
Abstract
This paper presents a comprehensive review of recent advancements in artificial intelligence (AI) techniques for the diagnosis of Alzheimer’s disease (AD). AD is a progressive neurodegenerative disorder whose complex pathology and heterogeneous clinical presentation make early and accurate diagnosis challenging. Traditional unimodal approaches that rely on a single source of information such as MRI, PET, cerebrospinal fluid biomarkers, or cognitive assessments often fail to capture the full spectrum of disease indicators. Recent advances in AI, particularly deep learning, has enabled multimodal data fusion that integrates complementary information from neuroimaging, biomarkers, genetics, and cognitive assessments. This review systematically analyzes state-of-the-art multimodal deep learning techniques for AD diagnosis, with an emphasis on fusion techniques that include early fusion, late fusion, adaptive weighting, and attention-based mechanisms. We compare and summarize findings across five major tasks: mild cognitive impairment (MCI) subtyping, prediction of MCI-to-AD conversion, early detection, risk prediction, and assessment of disease severity. Emerging trends highlight the increasing use of graph convolutional networks, transformer architectures, and attention mechanisms that enhance diagnostic accuracy and interpretability. Despite these advances, key challenges remain, including limited sample sizes, missing modalities, data heterogeneity, and insufficient clinical validation. This review concludes by emphasizing the need for scalable, explainable, and clinically deployable AI frameworks capable of supporting precision diagnosis and enabling personalized treatment planning for individuals affected by AD.