Advancing early detection of Alzheimer's disease: from established biomarkers to future innovations

Alzheimer's disease (AD) is a major form of dementia, imposes an escalating global burden on public health and social care. Pathological changes in AD occur years before clinical symptoms, defining a prolonged preclinical phase. Currently, there is no effective treatment exist for late-stage AD, making early diagnosis and intervention indispensable, in which biomarkers play a pivotal role. While cerebrospinal fluid (CSF) and positron emission tomography (PET) biomarkers improved AD diagnostic accuracy, their clinical application is limited by invasiveness, high costs. Advances in ultrasensitive immunoassay technologies has rendered peripheral blood biomarkers easily accessible, cost-effective, minimally invasive, and suitable for continuous monitoring, offering noval opportunities for AD screening, diagnosis and progression tracking. Additionally, non-invasive body fluids including saliva, urine and tears are being explored as potential biomarkers. Artificial intelligence (AI)-driven multimodal fusion, integrated with large-scale multimodal computing frameworks, significantly improve the accuracy of early AD diagnosis. Recent studies have identified a series of promising synaptic and neuronal biomarkers, such as neurofilament light chain (NFL), neurogranin (Ng), β-secretase 1 (BACE1) and synaptosomal-associated protein 25 (SNAP-25). Inflammatory mediators and genetic factors also contribute to AD pathological evaluation. This review systematically characterizes biomarker profiles of emerging non-invasive carriers (saliva, urine and tears). It further synthesizes cohort-derived quantitative diagnostic metrics (AUC, sensitivity, specificity) of multimodal AD biomarkers and conducts head-to-head comparison to clarify their respective strengths and translational limitations, aiming to provide evidence-based insights for optimizing early AD screening and diagnostic strategies.