Artificial Intelligence for Early Detection of Alzheimer's Disease: A Systematic Review of Opportunities, Challenges, and Future Prospects

Abstract

With aging populations, Alzheimer's disease (AD) affects >50 million and demands early or prodromal detection in EOAD and LOAD. Current diagnostics lack precision, reach, and affordability. AI enables scalable, multimodal, noninvasive screening. Review AI advances for early AD detection across neuroimaging, speech, wearables, EEG, and other biomarkers; emphasize multimodal integration, address practical and ethical issues, and recommend priorities. Following PRISMA 2020, we searched six databases. Inclusion: peer reviewed AI and AD diagnosis or management studies reporting empirical outcomes with ≥100 participants. We extracted precision, modality, limitations; quality via JBI; PROSPERO CRD42024512345. Out of 1,456 records, 41 high-quality studies remained. AI achieved 85–98.5% accuracy; multimodal models outperformed (AUC up to 0.98). Persistent issues include bias, interpretability, generalizability, and privacy and equity concerns. AI is reshaping early AD detection via multimodal explainable approaches; realizing requires diverse datasets, ethics, federated learning, collaboration, multisite longitudinal XAI, and equitable deployment.