Related Experiment Videos
Mapping the landscape of AI-driven digital twins in medical diagnosis: A scoping review on core technologies,
Shiying Shen1, Wenhao Qi2, Xin Liu2
1School of Nursing, Hangzhou Normal University, Hangzhou, China; Department of Neurology, The Affiliated Hospital of Hangzhou Normal University, Hangzhou, China; Zhejiang Philosophy and Social Science Laboratory for Research in Early Development and Childcare, Hangzhou Normal University, Hangzhou, China.
Objective:
This study aims to map the application landscape, core technical components, key challenges, and future directions of digital twin technology in medical diagnosis through a scoping review and bibliometric analysis.
Methods:
Following PRISMA-ScR and BIBLIO guidelines, we searched five databases from inception to July 16, 2025, identifying 64 eligible studies. We analyzed data on research landscape (national output, collaboration, funding, interdisciplinarity), technical components (data sources, modeling), model maturity, and application scenarios.
Results:
Global output was concentrated in the United States (19.4%), the United Kingdom (10.9%), and China (10.1%), which formed the core hubs of international collaboration. Funding was primarily from government (40.4%) and nonprofit organizations (26.6%). The field is AI-centric and notably interdisciplinary. Diagnostic digital twins commonly adopt hybrid modeling that combines physics-based simulation with data-driven analytics. Medical imaging was the dominant data source (70.3%), and deep learning served as the principal algorithmic driver. Overall technological maturity remained low: most models (97%) were at L2-Conception and L3-Operations, while systems with real-time, closed-loop feedback (L4) were rare. Current applications focus on automated lesion detection, individualized risk stratification, and dynamic monitoring/diagnostics, with promising accuracy in cardiovascular diseases, oncology, and neurologic disorders.
Conclusion:
Digital twins are a key enabling technology for predictive, personalized, and systemic precision diagnostics. Translation faces data scarcity, model computational burden, insufficient validation, limited interpretability, workflow integration, and regulatory gaps. Future work should prioritize multi-center data federations, efficient and trustworthy hybrid modeling, large-scale clinical validation, and adaptive regulatory frameworks to accelerate clinical adoption.
