Evolution of digital twins in precision health applications: a scoping review study

Jiang Bian, Yu Huang, Hao Fu Dai, Jie Xu, Ruoqi Wei, Leyang Sun, Yi Guo, Jingchuan GuoResearch Square 2024 · DOI

Abstract

An increasing amount of research is incorporating the concept of Digital twin (DT) in biomedical and health care applications. This scoping review aims to summarize existing research and identify gaps in the development and use of DTs in the health care domain. The focus of this study lies on summarizing: the different types of DTs, the techniques employed in DT development, the DT applications in health care, and the data resources used for creating DTs. We identified fifty studies, which mainly focused on creating organ- (n=15) and patient-specific twins (n=30). The research predominantly centers on cardiology, endocrinology, orthopedics, and infectious diseases. Only a few studies used real-world datasets for developing their DTs. However, there remain unresolved questions and promising directions that require further exploration. This review provides valuable reference material and insights for researchers on DTs in health care and highlights gaps and unmet needs in this field.

Introduction: The Identity Crisis of Health Digital Twins

The concept of the Digital Twin (DT)—a virtual model dynamically linked to a physical entity—is migrating from aerospace and manufacturing into precision medicine. However, the healthcare domain lacks a clear taxonomy and a unified understanding of what actually constitutes a Health Digital Twin (HDT).

Researchers often conflate DTs with older simulation methods like Microsimulation (MSM) or Agent-Based Modeling (ABM). This paper addresses this "identity crisis" by clarifying the distinctions between these methods and providing a comprehensive overview of the current state-of-the-art in HDT research, identifying unmet needs in data integration, model fairness, and real-world implementation.

Comparison of Digital Twin, Microsimulation, and Agent-based Simulation concepts and applications.

Methodology

The authors performed a two-phase scoping review following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. The first phase involved analyzing 25 existing review articles to define the scope and distinguish DTs from MSM and ABM. The second phase consisted of a detailed analysis of 50 individual primary studies published between 2013 and 2023.

The two-phase selection process for the scoping review, resulting in 25 review articles and 50 primary studies for analysis.

Studies were categorized based on the physical entity represented (e.g., organ vs. patient), core modeling techniques (AI/ML vs. mathematical), and clinical application domains using the Phecode system.

Data Sources and Modeling Techniques

The study identifies three primary categories of data used to build HDTs:

  • Prior Clinical Studies (58%): Data from clinical trials and observational cohorts (e.g., ADNI and the Human Connectome Project).
  • Real-World Data (22%): Electronic Health Records (EHRs) and administrative claims (e.g., MIMIC-III).
  • Simulation-Generated Data (20%): Synthetic data used primarily for model validation.

A significant shift toward AI/ML (58% of studies) over traditional mechanistic mathematical models (36%) was observed. Deep learning architectures like CNNs and RNNs are becoming the standard "engines" for HDTs, though Reinforcement Learning is emerging for treatment optimization.

Key Insights: Hype vs. Reality

A critical finding is the "clout-chasing" phenomenon: many researchers use the "Digital Twin" label for work more accurately described as traditional simulation. A true DT requires a "live connection" with bidirectional data exchange—a feature almost entirely absent in current literature. Most models remain "static" snapshots rather than evolving "living" models.

In terms of scale, research focuses on Patient-based twins (60%) for holistic monitoring and Organ-based twins (30%), with the heart being the most frequently modeled organ.

Clinical Applications and Results

Cardiovascular medicine is the leading application area, followed by oncology, endocrine/metabolic disorders (specifically Diabetes), and infectious diseases (COVID-19). Organ-based DTs are primarily used for forecasting treatment responses, while patient-based DTs are aimed at long-term health monitoring and "what-if" scenario testing for chronic disease management.

Challenges and Future Directions

To move from technical curiosity to clinical utility, several hurdles must be overcome:

  • Interpretability: The shift toward deep learning necessitates Explainable AI (XAI) for clinical adoption.
  • Bias and Fairness: There is a glaring lack of attention to algorithmic fairness; training on biased real-world data risks exacerbating health disparities.
  • Multimodal Integration: Future HDTs should utilize Multimodal Foundation Models to integrate genomics, imaging, and social determinants of health.
  • Human-AI Teaming: HDTs must be co-created with clinicians and patients to ensure they solve real-world clinical problems.

Conclusion

While Health Digital Twins hold immense promise for precision medicine, the field is currently characterized by a gap between conceptual definitions and practical implementations. Moving forward requires establishing rigorous standards for bidirectional data exchange and ensuring models are fair, interpretable, and clinically relevant.

Summary updated

Abstract

An increasing amount of research is incorporating the concept of Digital twin (DT) in biomedical and health care applications. This scoping review aims to summarize existing research and identify gaps in the development and use of DTs in the health care domain. The focus of this study lies on summarizing: the different types of DTs, the techniques employed in DT development, the DT applications in health care, and the data resources used for creating DTs. We identified fifty studies, which mainly focused on creating organ- (n=15) and patient-specific twins (n=30). The research predominantly centers on cardiology, endocrinology, orthopedics, and infectious diseases. Only a few studies used real-world datasets for developing their DTs. However, there remain unresolved questions and promising directions that require further exploration. This review provides valuable reference material and insights for researchers on DTs in health care and highlights gaps and unmet needs in this field.
BibTeX
@inproceedings{bian2024evolution,
  author = {Jiang Bian and Yu Huang and Hao Fu Dai and Jie Xu and Ruoqi Wei and Leyang Sun and Yi Guo and Jingchuan Guo},
  title = {{Evolution of digital twins in precision health applications: a scoping review study}},
  booktitle = {Research Square},
  year = {2024},
  doi = {10.21203/rs.3.rs-4612942/v1},
}
Figure