Blood biomarkers and artificial intelligence: toward precision medicine in Alzheimer’s disease
TL;DR
This paper synthesizes recent advancements in the application of machine learning on blood biomarker analysis, evaluating their diagnostic efficacy and clinical translational potential and proposes a scalable “Step-Care” framework that transitions from single-biomarker community screening to precise multimodal fusion, optimizing clinical trial stratification and personalized workflows.
Abstract
The integration of blood-based biomarkers and artificial intelligence (AI) is revolutionizing the diagnostic landscape of Alzheimer’s disease (AD). In this paper, we synthesize recent advancements in the application of machine learning (ML) on blood biomarker analysis, evaluating their diagnostic efficacy and clinical translational potential. Current evidence identifies plasma p-tau217 as the best individual biomarker for primary screening. However, multiple biomarker signatures (such as A/T/N) combined with ML algorithms (like XGBoost) greatly improve diagnostic accuracy for mild cognitive impairment. Notably, predictive performance is maximized with multimodal fusion, which integrates data from blood, neuroimaging, and genetics, achieving an area under the curve of up to 0.94. To mitigate the black-box nature of these models, explainable AI frameworks like SHapley Additive exPlanations (SHAP) are essential for establishing clinical interpretability. Ultimately, we propose a scalable “Step-Care” framework that transitions from single-biomarker community screening to precise multimodal fusion, optimizing clinical trial stratification and personalized workflows.