Research led by Natalia Trayanova, PhD, director of the Alliance for Cardiovascular Diagnostic and Treatment Innovation at Johns Hopkins University is testing AI’s ability to not only detect cardiac arrhythmias, but to recommend treatment.
The technology, known as diffeomorphic mapping operator learning, or DIMON, is able to predict cardiac arrhythmias by analyzing a patient’s digital heart twin. This process used to take days for Dr. Trayanova and her team to complete. With DIMON, the process now only takes seconds, according to a Dec. 9 news release from Johns Hopkins University.
DIMON’s framework is now being used to solve complex engineering problems within other research fields, the release said.
Dr. Trayanova, who is also director of AI research in health and medicine at Johns Hopkins’ Data Science and AI Institute, spoke to Becker’s about her research and how it can be integrated into clinical cardiovascular care.
Editor’s note: Responses have been lightly edited for clarity and length.
Question: Can you provide a brief explanation of the technology?
Dr. Natalia Trayanova: We take cardiac imaging and data from the patient to create a personalized digital heart twin. We then use different simulations and scenarios on the digital twin to see what treatment provides the best outcome. That information can then be used by a physician to propose a treatment plan.
Q: What are the next steps for bringing this technology into routine clinical practice?
NT: To make this process part of clinical workflow, it has to be made executable on a desktop computer. For a hospital to embrace it, it would require an investment in deploying the technology at scale.
AI technology has to be deployed in such a way that the physicians can have trust in it. They may need to interact with it for a while to be convinced, so interpretability is very important.
Using the technology through a personalized digital heart twin is a mechanistic, tangible guide to provide decision support. Physicians can look at it, zoom in on it, see why certain things are happening, or why not. But then physicians also need to have a way to explain the AI findings — the specific features or characteristics identified by algorithm — to their patients, and be able to communicate why treatment decisions are being made.
