Ethical digital twins can help make India disease poor
Premium- It is expected that digital twins will be an integral part of future medicine
Back
PremiumImagine the digital representation of a physical object such as an aircraft engine; its virtual replica or a “digital twin" could be used to predict and avert catastrophic failures. Science fiction? No, it is already happening. Fuelled by large amounts of data, the internet of things (IoT), and artificial intelligence (AI), remarkable progress has been made in our ability to generate models that can predict future events in near real-time for a variety of critical systems. This is one of the top 10 future technology trends in engineering that is also coming to MedTech.
However, living systems are far harder to represent digitally compared to even the most complex inanimate object. The challenges are both technical and ethical.
At a technical level, the issue is one of sufficient data and understanding. Today, at most, one can model specific organs or functions. For example, using advanced implanted pacemakers and other sensors, the physiology of a beating heart can be represented as a digital twin, on which interventions can be planned and tested.
Cardiologists would much rather learn from a failure that sends the digital doppelganger into a fatal arrhythmia than the real one, thus allowing better outcomes for actual patients. Similar efforts for creating limited digital twins are ongoing for other organs, response to drugs, and the effect of different diets, among other things. Progress has been slow but steady, with an expectation that this will be an integral part of future medicine.
Things are significantly more complex, technically and ethically, when it comes to measuring and representing innate and heritable aspects of human beings—their genetic code. At one level, one would argue that without knowing the fundamental blueprint of the individual, one cannot construct an adequate digital twin.
Yet, it is equally true that excess information that is not actionable is dangerous and unlikely to lead to patient benefit. Many variations in the genome inform only of distant risks, such as those of breast cancer or Alzheimer’s disease, and that too in terms of widely ranged odds. While a few mutations do reliably inform of genetic disease, the fact remains that actionable disease risks remain mostly unchanged before and after genome sequencing, except for families with known heritable genetic diseases. In many situations, it is not clear what one should do, even with reliable genetic risk information. Consider the case of a young girl found to have a BRCA2 genetic mutation that increases the risk for breast or ovarian cancer later in life. What exactly is to be done?
Clearly, there is little benefit to telling her immediately. If she is to be informed later, then by whom and when? How would she get reminders for mammograms or other breast cancer screening, and who would examine her health data to look for any signs of early cancer? There are many challenges here, all leading eventually to the same answer – ethical design of digital twins for healthcare, such that approved AI-enabled digital health systems are appropriately monitoring the digital doppelganger.
The foremost requirement of an ethical digital twin for healthcare is institutional anchoring with lean but balanced governance. Massive amounts of private data with potential for misuse will go into such twins, whether for organs or genomes or just routinely collected data from wearables, lab tests, etc.
The privacy rights of individuals must be protected while ensuring that maximum learning is enabled from the data without compromising the data. One such way is federated learning, where the data is secure, and only algorithms and parameters move. We are currently working to determine the feasibility of such an approach with routine electronic health records.
While we await the next stage of deep digital representation of living systems, digging deeper into the digital health resources we already have may be useful. Based on the simple idea that the course of an illness in a new patient is likely to match with previous patients, Stanford Hospital now provides specialized consults involving searching a library of electronic patient records for medical answers to meet the needs of unusual patients.
This has been shown to be a valuable approach and used to guide therapy in complex cases. Currently, Indian hospitals are disease-rich and data-poor. One hopes that the national digital health mission will be the dawn of a new age where we are data-rich and disease poor.
Anurag Agrawal is the dean, bio-Sciences and health research at Ashoka University.
Catch all the Business News, Market News, Breaking News Events and Latest News Updates on Live Mint.
Download The Mint News App to get Daily Market Updates.
More
Less
