AC LITERATURE

Dr. Anthony Chang

“A digital twin (in health) should be individualized, interconnected, interactive, informative, and impactful (the 5 “i”s).”
- Evangelia Katsoulakis, Digital Twin Researcher

Clinicians often face challenging cases without enough information and knowledge (and especially without wisdom) in the clinical setting, and must rely on their own limited experience to manage these patients. In this era of big data and artificial intelligence, it remains difficult to know what the ideal management is for each individual patient (precision medicine). The complex dynamical systems of humans need a smarter strategy beyond machine and deep learning.  

Digital twins are virtual models or representations of real world entities characterized by three components: 1) the physical entity (cells, organs, diseases, patients, and even populations), 2) its virtual replica, and 3) a connection where real-time data can be exchanged continuously in the form of digital threads. The origin of the digital twin was the Apollo 13 mission when a “twin” was assembled to engender real-time engineering solutions to bring the astronauts safely home, but some feel that it was human ingenuity with the astronauts and engineers that really was the most contributory (perhaps a bit of human hubris?). Since then, digital twins have been used in manufacturing, architecture, military operations, and a myriad other domains. While early models of digital twins have been relatively static, more recent renditions of digital twins are more functional with dynamic capabilities via digital threads and are now sometimes called shadow twins. This latter modern version of a “living” digital twin has a real time and continuous connection to enable optimization of simulation (closed loop optimization). 

In healthcare AI, digital twins can be made from multiple sources of data such as electronic health records, imaging studies, genomic information, wearable device data, social determinants of health, population data, etc. This health virtual twin allows for dynamic simulation of diagnosis, treatment, and monitoring for any medical condition with a trajectory towards maximal health for potentially the entire lifetime. Among the many domains that have leveraged digital twins include: precision medicine, cancer care, cardiovascular disease, chronic disease management, hospital management, facility design, workflow development, and individualized training. The utility of digital twins in healthcare can be from not only the microscopic to the macroscopic but also from birth to death, therefore its impact can range from precision medicine to public health. 
In the near future, a convergence of generative AI, cognitive computing, internet of everything (IoT with embedded AI), and deep reinforcement learning will exponentially increase the capabilities of digital twins to become “intelligent” twins. This intelligent twin effort will need a trans disciplinary approach of experts in medicine, engineering, computer science, data science, physiologists, basic scientists, experts in ethics, regulation, and law, and most importantly, patients and families. To add an interesting part of the ecosystem in the future, multi-agent AI can become a valuable asset in this growing portfolio of AI tools for the future clinicians to attain the Quintuple Aim.

These insights and discussions on digital twins in AI in healthcare and how we can successfully educate ourselves in these topics in AI will be discussed at the in-person AIMed24 meeting scheduled for November 17-19, 2024 at the sublime Caribe Royale resort in Orlando, Florida. 

See you there!

ACC