Digital Twins for Predictive Modelling of Thrombosis and Stroke Risk: Current Approaches and Future Directions

Abstract

Thrombosis drives substantial global mortality across atrial fibrillation, venous thromboembolism, and atherosclerosis. However, clinical scores treat risk as a static variable and omit evolving comorbidities, functional biomarkers, anatomy, and treatment exposure, leading to misclassification and preventable events. This statement advances a unified scientific agenda for patient-specific digital twins that dynamically integrate multimodal longitudinal data with mechanistic insight to predict thrombogenesis risks. We position these digital twins as hybrid models anchored in physics and data-driven algorithms that can simulate disease progression and therapy. The goal of this approach is to refine stroke and bleeding estimation beyond current clinical rules. Continuous updating from imaging data, laboratory test results, wearables, and electronic health records supports dynamic risk trajectories and adaptive care pathways, facilitating continuous risk reassessment. This statement analyzes gaps in data quality, calibration, validation, and uncertainty quantification that presently limit the clinical translation of this technology. Research priorities are then proposed for multiscale thrombosis modelling, physics-informed learning, probabilistic forecasting, and regulatory-compliant data stewardship. Finally, we outline translation to in silico trials, regulatory alignment, and hospital workflows that link predictions to decisions. By articulating shared challenges across thrombosis-driven diseases and reframing risk as a time-varying measurable quantity, this statement lays a foundation for developing digital twin approaches that support a shift from population heuristics towards precise, timely thrombosis care. These advances are essential for translating digital twin technology from research to clinical practice, enabling dynamic risk prediction and personalized anticoagulation therapy.