Digital twins are virtual representations of an object or system that spans its lifecycle, is updated from real-time data, and use simulation, machine learning and reasoning to help decision-making (IBM). In most cases, this helps data scientists understand how products are operating in production environments and anticipate how they may behave overtime. But what happens when a digital twin is that of a human being?
By using digital twins to model a person, you can use technologies like natural language processing (NLP) to better understand data and uncover other useful insights that will help improve use cases from customer experience to patient care. Today, we’re simply generating more data than ever before. Digital twins can be useful in synthesizing this information to provide actionable insights.
As such, there are few fields digital twins can be more helpful in than healthcare. Take a visit to your primary care physician, for example. They will have a baseline understanding of you — your history, medications you take, allergies, and other factors. If you then go to see a specialist, they may ask many of the same repetitive questions, and remake inferences and deductions that have been done before. But beyond convenience and time savings, digital clones can substantially help with accuracy.
Having a good virtual replica of a patient enables medical professionals to dig down specific medications, health conditions, and even social determinants of health that may impact care. Greater detail and context enables providers to make better clinical decisions, and it’s all being done behind the scenes, thanks to advances in artificial intelligence (AI) and machine learning (ML).
Digital Twins in Production
Digital clones or digital twins can greatly benefit the healthcare system, and we’re already starting to see them in use. Kaiser Permanente uses digital twins through a system that improves patient flow within a hospital. It achieves this by combining structured and unstructured data to build a more complete view of each patient to anticipate what their needs will be at the hospital. In another instance, Roche uses digital twins to help securely integrate and display relevant aggregated data about cancer patients into a single, holistic patient timeline.
Digital twins are already at work in some of the largest healthcare organizations in the world, but their potential doesn’t stop with the existing use cases. There are many other applications for digital twins at play, and they span from practical everyday use to functions that sound more like science fiction than reality. Here are some additional areas digital twins can be particularly useful in healthcare:
Summarizing Patient Data: Providers are experiencing information overload with the amount of data in today’s healthcare system. From electronic health records (EHRs) to doctor’s notes to diagnostic imaging, it can be a challenge to connect the disparate data — structured tables, unstructured text, medical images, sensors and more — associated with an individual patient. Consider a patient with a cancerous tumor along with other underlying conditions. Typically, oncologists and other specialists will meet to determine the next steps in treatment, whether it be surgery, medication, or another protocol. Integrating all this data into a unified, relevant, and summarized timeline can be done using a combination of natural language processing (NLP), computer vision (CV), and knowledge graph (KG) techniques today.
Accelerating Precision Medicine: Precision medicine is mostly applied in the areas of cardiology and oncology, dealing with serious conditions, as cancer and heart disease. Sometimes, instead of recommending an aggressive treatment like chemotherapy, it’s important to see if a patient has certain genomic biomarkers that can inform doctors if another approach may work better for that patient. Genetic profiling is useful to uncover these insights, helping doctors better understand a given patient’s tumor, labs, genomics, history, and other pertinent details to reach an optimal decision. As a result, the clinician can provide a more personalized approach to care. However, to achieve this, you need to aggregate much more information about the patient. By building a digital twin, you can compare an individual to other patients – similar in clinically important ways – to see if there are genomic similarities and how certain treatments have impacted them.
Process Improvement: Improving organizational performance, thereby improving patient outcomes or population health, requires a high level of specificity. For example, if your goal is to reduce the length of a patient’s hospital stay, it’s imperative to understand many other factors about their condition. Through structured data, you can find information, like whether the patient has a chronic condition and what medications they were taking, or whether or not they have insurance. But some of the considerations that really matter in terms of the duration of the patient’s hospital stay — how they are eating, feeling, sleeping, coping, moving, etc. — can only be found in free-text data. Creating a digital twin to anticipate patient needs and the length of their stay can be very valuable.
What’s Next for Digital Twins
Some medical devices have the capabilities of producing digital twins of specific organs or conditions so doctors can better diagnose them. Areas like NLP can be a great help here if you have a patient with a chronic condition (Asthma, COPD, mental health issues, and others). For acute issues – especially in oncology, cardiology, and psychiatry – digital twins can offer a higher level of detail. For example, creating the digital twin of a patient’s heart enables a doctor to see exactly what’s going on — whether there is scarring from previous surgeries or an abnormality that needs to be inspected further — and make better decisions before an operation, rather than during. This can mean a world of difference for patient outcomes.
We’ll start to see more advanced use cases for digital twins in the coming years. But to truly live up to the hype, it’s crucial that we move beyond simply collecting and analyzing only structured data. Recent advances in deep learning and transfer learning have made it possible to extract information from imaging and free-text data, serving as the connective tissue between what can be found in EHRs and other information, like radiology images and medical documents of all types. Only then can we begin to construct a meaningful digital twin to uncover useful insights that will help improve hospital operations and patient care.
About David Talby
David Talby, Ph.D., MBA, is the CTO of John Snow Labs, the AI and NLP for healthcare companies provide state-of-the-art software, models, and data to help healthcare and life science organizations put AI to good use. He has spent his career making AI, big data and data scientists solve real-world problems in healthcare, life science and related fields.