The DigiTwin Study: Where AI, Wearables, and Personalized Medicine Converge
What if researchers could combine imaging, genetics, laboratory testing, wearable data, and environmental measurements to create a dynamic model of one person’s health? An ambitious study is now exploring that possibility by collecting comprehensive health data from 1,000 participants over time.
In an earlier BLS article, we examined the emerging use of digital twins in medical research, surgical planning, and drug development. A medical digital twin is a computational model of an organ, biological system, or individual that can be updated with real-world data to simulate conditions, examine change over time, or generate predictions.
Researchers at the BioMedical Engineering and Imaging Institute at the Icahn School of Medicine at Mount Sinai are now applying the concept at the whole-person level. The DigiTwin Study is a longitudinal, observational study designed to collect and integrate multimodal health data. Researchers intend to use the resulting dataset to study health trajectories, including biological aging and the development of cardiovascular disease. The study is an early research step intended to provide the data foundation for future digital-twin models.
Building a Whole-Person Model
The broader DigiTwin Project reflects an effort to move healthcare toward earlier risk identification and more personalized prevention. Its longer-term vision is to combine longitudinal health monitoring with artificial intelligence to identify changes that may precede conventional clinical findings and potentially extend “healthspan,” the period of life spent in good health.
The current study does not provide diagnoses, treatment, individualized medical advice, or clinical recommendations. It also does not evaluate the safety or effectiveness of an AI-based medical tool. Mount Sinai identifies the study as exploratory and observational and states that the information collected is for research purposes.
An Unusually Broad Dataset
The study will collect information over 24 months for each participant, with in-person testing at baseline and follow-up visits after one and two years. The full 1,000-person cohort is expected to be assembled over approximately three to five years.
The study combines initial and periodic measurements, such as whole-body MRI, retinal imaging, and multiomics profiling, with information collected more continuously through wearable and at-home devices. That combination may allow researchers to examine a participant’s broader biological profile while also tracking changes that may occur between clinical visits.
Multiomics is a particularly notable component of the study because it brings together several layers of biological information. As discussed in another recent BLS article, analyzing those layers together may provide a more complete picture of the biological processes involved in health and disease.
The breadth of the dataset is central to the project because aging and disease do not necessarily remain confined to one organ system. Combining molecular, physiological, imaging, behavioral, and environmental information may help researchers identify relationships that would be missed when each source is studied separately. That breadth also creates practical difficulties. Imaging, laboratory, wearable, environmental, and participant-reported data may use different formats, measurement methods, and collection intervals. Missing information, participant adherence, device variability, and differences in sample processing can affect the reliability of later analysis.
The Role of AI and Machine Learning
Researchers plan to use AI and machine learning to analyze population-health datasets and the eventual DigiTwin cohort, looking for patterns in biological aging and health trajectories that may precede conventional symptoms or clinical findings. Models may identify associations, group participants with similar characteristics, and examine how changes in sleep, exercise, diet, environmental exposure, or other factors correspond with biological measurements.
More data do not automatically produce better evidence. Missing data, inconsistent acquisition methods, demographic differences, and bias can limit reproducibility and generalizability. Model performance will depend on data quality, population representation, external validation, and a clearly defined intended use.
Researchers must also distinguish prediction from causation. A model may detect a pattern that precedes disease without showing that the pattern causes the disease or that changing it will improve an outcome. Before model-generated insights can support diagnosis, treatment, or individualized prevention, their analytical performance and clinical relevance must be established for those uses.
A Public Illustration
A recent TODAY report introduced the research to a broader audience. The segment followed a reporter’s completion of in-person testing and at-home monitoring and showed how the data collected were used to create a digital twin and model the potential effects of lifestyle changes such as sleep, exercise, and diet.
This snapshot of the DigiTwin Project in action illustrates how digital models might be used to identify and analyze potential health risks before physical symptoms are recognized. The report makes the concept tangible, but it remains an illustration of the technology rather than evidence of improved clinical outcomes.
The report estimated that the testing required to create the reporter’s digital twin would cost approximately $5,000 to $6,000 per person. That estimate underscores a practical challenge: even if whole-person models prove clinically useful, their broader adoption may depend on whether the underlying testing, data integration, and analysis can be delivered at a sustainable cost.
Moving from a research demonstration to dependable clinical use will require careful management of the cost, data, technology, validation, and governance issues surrounding the model.
Privacy, Governance, and Operational Risk
For healthcare, biotechnology, medical-device, and digital-health organizations, the DigiTwin Study illustrates the convergence of clinical research, software, AI, diagnostics, connected devices, cloud services, and sensitive health information. These activities create privacy, cybersecurity, consent, and data-governance considerations.
Whole-person digital models require organizations to combine substantial amounts of sensitive information, potentially including genetic data, medical imaging, laboratory results, wearable measurements, lifestyle information, and environmental data. As discussed in a prior BLS article on genetic testing and privacy, genetic information raises particular concerns because it may remain connected to an individual after conventional identifiers are removed and reveal information relevant to biological relatives. Wearable and environmental data may expose additional information about a person’s daily activities, health, and home environment.
Regulatory analysis will depend partly on intended use and product claims. A research model that investigates associations presents a different profile from software marketed to diagnose disease or recommend treatment. As discussed in another BLS article on the FDA’s 2026 guidance for low-risk digital-health products, the distinction between general-wellness products, clinical decision-support software, and regulated medical devices can turn on a product’s functions, claims, and role in clinical decision-making.
The combination of advanced technology and sensitive data may create overlapping contractual, regulatory, privacy, cybersecurity, technology errors and omissions, product liability, and clinical-research exposures. The liability and insurance analysis will depend on the organization’s role, contractual obligations, applicable law, the alleged injury or failure, and the wording of its insurance policies.
What Comes Next
The DigiTwin Study offers a useful view of where personalized health research may be heading. Researchers are attempting to understand how imaging, laboratory results, genetics, physiology, behavior, and environmental exposures interact across the same person and over time.
The current study is building the data foundation for that work. Whether the broader vision succeeds will depend on representative data, scientific and clinical validation, responsible governance, economic scalability, and evidence that model-generated insights improve decisions and outcomes. For now, the DigiTwin Study is significant because of the scale and breadth of the questions it is designed to investigate. .
Authored by Phillip Skaggs, J.D., Berkley Life Sciences, Vice President, Chief Legal and Regulatory Affairs Officer
This post is for general informational purposes only and is not intended as legal or other professional advice.