Multimodal approaches and AI-driven innovations in dementia diagnosis: a systematic review

• Published in: Discover Artificial Intelligence Vol. 5; no. 1; pp. 96 - 32
• Main Authors: Wahul, Revati M., Ambadekar, Sarita, Dhanvijay, Deepesh M., Dhanvijay, Mrinai M., Dudhedia, Manisha A., Gaikwad, Varsha, Kanawade, Bhavana, Pansare, J. R., Bodkhe, Balaji, Gawande, S. H.
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2025; Springer Nature B.V; Springer
• Subjects: Algorithms; Alzheimer's disease; Alzheimer’s disease (AD); Artificial Intelligence; Brain research; Cognitive ability; Computer Science; Data analysis; Deep learning (DL); Dementia; Dementia diagnosis; Disease; Engineering; Frontotemporal dementia (FTD); Innovations; Machine learning; Machine learning (ML); Medical imaging; Memory; Multimodal data Integration; Mutation; Neuroimaging; Paradigms; Proteins; Public health; Review; Systematic review; Alzheimer’s disease (AD); Explainable AI (XAI); Machine learning (ML); Multimodal data Integration; Deep learning (DL); Dementia diagnosis; Frontotemporal dementia (FTD)
• ISSN: 2731-0809; 2731-0809
• DOI: 10.1007/s44163-025-00358-x

Neurodegenerative disorders, such as dementia, present some of the most pressing challenges in the field of medicine today. By causing progressive cognitive and functional decline, Alzheimer’s Disease (AD) and Frontotemporal Dementia (FTD) subtypes are an essential area for urgently needed work. As a systematic literature review, this paper outlines the intricacies of dementia’s pathophysiology of dementia by addressing the complexity of dementia as a construct, how different types of clinical paths can happen, how neuronal atrophy occurs along different cerebral domains, and when the critical diagnostic thresholds are met. A complete review of peer-reviewed papers over the past ten years, with a focus on Machine Learning, Deep Learning, and multimodal fusion approaches to enhance diagnostic and therapeutic precision will focus on neuroimaging biomarkers, EEG-based cognitive profiles, digital phenotyping, and wearable sensor analytics. This survey will compare the study’s algorithms or frameworks on sensitivity, specificity, interpretability, computational efficiency, and clinical transnationality concerning early detection and monitoring progression. Though AI methods are having a continuing rapid surge in progress, the issues of model transparency and generalizability are still lacking, thus meaning the need for XAI. This work builds a multi-disciplinary data agnostic approach for building stronger patient-centered models that can bring together genomics, imaging, behavior and contextual features in the task-driven processes. Overall, this literature survey’s objective is to shine a light on the multi-faceted pathway towards precision-driven, AI augmented dementia care—and ultimately to change the management of neurodegenerative disease by synthesizing current developments and highlighting their shortcomings.