How digital copies make medicine more personal

University of Illinois Chicago researchers are seeing double.

Lu Cheng, a computer scientist in the College of Engineering, studies digital twins: computer-dwelling duplicates of real-world systems.

Cheng and colleagues from UIC and Pennsylvania State University received $900,000 from the National Science Foundation to create digital twins of patients with Alzheimer's disease. Replicating Alzheimer's patients' body measurements, blood composition and internal organs in the digital dimension can help clinicians diagnose, treat and predict the disease's effects, Cheng said.

Digital twins aren't just for health care. UIC researchers also use them to study neutrinos - tiny, chargeless particles that hold the secrets of space and matter - and some are investigating the AI framework that makes digital twins possible.

In a conversation with UIC today, Cheng elaborates on digital twins and their role in her research.

What is a digital twin?

A digital twin is exactly what it sounds like. It's a digital model of a real-world physical system - in our case, an organ or even a whole person.

How are digital twins used in health care?

Digital twins in health care can take patients' clinical data - such as medical records, imaging and wearables - as input to continuously update. We can predict disease progression, plan surgeries, optimize hospital operations as well as examine, understand and intervene with treatments without touching the real person. Testing "what-if" scenarios with digital twins can help clinicians make safer decisions.

What does a digital twin look like?

A digital twin isn't a single object. Think of it as a living computational model: a structured collection of numbers, relationships and rules that update over time. Sometimes it looks like tables of measurements, sometimes like a network and sometimes like simulated "what-if" outcomes. The key is that it behaves like the patient, not that it visually resembles one.

What kind of data informs a digital twin?

Unlike computer simulations, digital twins can co-evolve alongside patients in real time. Data from biosensors, heart monitors, smartphones and smart watches are constantly updating the digital twin, as well as doctors' notes and clinical measurements like blood samples and diagnostic imaging. That means we can create a digital version of every patient, and they'd all be unique - just like people.

What would using digital twins look like in a hospital setting?

Digital twins are a collaboration between the patient, who contributes data, and the doctor or clinician who studies their case. A doctor can model how a disease might progress in someone or affect them specifically or simulate a treatment plan.

Your research focuses on patients with Alzheimer's disease. Why are you starting here?

Alzheimer's disease is a natural starting point for us because it's progressive, slow-moving and complex - we still don't know what causes it. Patients with similar symptoms can follow very different trajectories, which makes it an ideal testbed for digital twins that model individualized disease progression rather than population averages.

What might a digital twin help clinicians understand about a patient with Alzheimer's?

A digital twin could help answer a question like if we intervene now, like changing medication, sleep or vascular risk factors, how might this patient's cognitive decline change over the next year compared to doing nothing? It's about comparing possible futures and understanding the disease's progression, not simply predicting one outcome.

Using digital twins in a clinic is still far off. Where are you now in the research process?

Right now, we're in the model-building and validation stage: learning how to combine scientific literature, clinical data, biomarkers and causal structure into reliable models. Before we can integrate digital twins into clinical workflows, we must build trust through rigorous testing and validation. This is a multiyear process.

Part of your research focus is responsible AI. What does responsibility mean in this context, and why is it important to consider?

Responsible AI means the models are transparent, uncertainty-aware, fair, reliable and decision-supporting, not decision-making. In medicine, errors have real consequences, so clinicians need to know when the model is or is not confident, fair and why. Responsibility is what turns AI from a risk into a tool clinicians can trust.