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

Alzheimer's disease is the leading cause of dementia worldwide, and its diagnosis in the early stages remains a significant clinical challenge, particularly because of the limitations of conventional methods regarding early sensitivity, cost, and availability. In this context, digital biomarkers combined with artificial intelligence have been investigated as noninvasive alternatives for identifying subtle cognitive changes. This study aimed to evaluate the diagnostic performance of artificial intelligence-based digital tools, including analysis of speech, writing, movement, and facial expressions, in the early detection of Alzheimer's disease. A systematic review of diagnostic accuracy studies was conducted in accordance with the PRISMA 2020 guidelines, with searches in the National Library of Medicine (United States), Virtual Health Library, Scientific Electronic Library Online, and Cochrane Library databases, over the period from 2020 to 2025. Of the 1040 records identified, 12 primary studies were included, and their methodological quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. Because of the heterogeneity of study designs and performance metrics, a narrative synthesis without meta-analysis was chosen, with the protocol previously registered in the International Prospective Register of Systematic Reviews (CRD420251248092). The results showed diagnostic accuracies ranging from 75.2 to 96.8%, with better performance observed in multimodal models and natural language-based approaches, frequently showing sensitivity and specificity above 90%, indicating the significant potential of these tools for early detection of the disease. It was concluded that artificial intelligence-based digital tools show promising performance as complementary strategies in the early diagnosis of the disease. However, their incorporation into clinical practice still requires multicenter validation, methodological standardization, and attention to ethical and contextual aspects.