Artificial intelligence and digital biomarkers in the early detection of Alzheimer's Disease: a systematic review

Abstract

Alzheimer’s disease (AD) is the leading cause of dementia worldwide, and early diagnosis remains an important clinical challenge, mainly due to the sensitivity, cost, and availability limitations of conventional diagnostic methods. In this context, artificial intelligence (AI)-based digital biomarkers have been investigated as noninvasive alternatives for the identification of subtle cognitive changes. This study aimed to assess the diagnostic performance of AI-driven digital tools, including speech, handwriting, movement, and facial expression analyses, in the early detection of AD. A systematic review of diagnostic accuracy studies was conducted in accordance with the PRISMA 2020 guidelines. Searches were performed in PubMed, Biblioteca Virtual em Saúde (BVS), SciELO, and the Cochrane Library between 2020 and 2025. Out of 1040 records identified, 12 primary studies met the eligibility criteria and were included. Methodological quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. Due to the heterogeneity in study design and performance metrics, a narrative synthesis was undertaken without meta-analysis. The review protocol was prospectively registered in the PROSPERO database (CRD420251248092). Diagnostic accuracy ranged from 75.2 to 96.8%. The highest performance was observed in multimodal models and natural language-based approaches, which frequently achieved sensitivity and specificity above 90%. These results highlight the considerable potential of these tools for early detection of AD. In conclusion, show promise as complementary strategies for the early diagnosis of Alzheimer’s disease. However, their integration into clinical practice still requires multicenter validation and standardization of analytical protocols to ensure their reliability and applicability across different populations and health care settings.