Improving heart transplant process with artificial intelligence

by Ryn Thorn 

Around 250,000 heart failure patients in the U.S. could benefit from heart transplants. For these individuals with end stage heart failure, a heart transplant is the best possible treatment, and yet only around 4,000 heart transplants are performed in the U.S. each year.

Arman Kilic, M.D., surgical director of MUSC’s Heart Failure and Heart Transplant Program and director of the Harvey and Marcia Schiller Surgical Innovation Center, has been awarded $1.9M for his proposed work to help shrink this chasm between transplant demand and transplant supply.

Kilic and his team, including Angel Jordan University Professor of Computer Science at Carnegie Mellon University Tuomas Sandholm, Ph.D., plan to use AI and machine learning to help optimize heart transplantation, hopefully saving lives in the process.

“We're aiming to change how organs are allocated,” said Kilic. “And once they are allocated, to optimize the decision making that's involved in accepting or rejecting those organs.”

Current state of heart transplantation

In addition to difficulties in filling heart transplant needs, transplant centers across the country evaluate donor hearts for potential recipients in different ways with varying outcomes.

According to Kilic’s team, donor non-utilization rates – when donor hearts are not used and discarded – are as high as 70-80%. This system breakdown may be due to the lack of adequate tools and time for transplant centers to efficiently match organs to recipients.

Existing methods to help transplant care providers match donor organs to those in need are possibly too simplistic and not used consistently from transplant center to transplant center.

There is a need for a stronger tool that can provide more consistent and efficient help with transplant decision making. Enter Kilic and his team.

Better transplants

Kilic’s approach to creating a better heart transplant system is threefold: 1) use machine learning to develop risk models that can aid in decision making, 2) using artificial intelligence to develop a better allocation policy for heart transplants, and 3) test and refine these methods by evaluating prototypes with end users.

Using current methods, medical care teams decide which patients will receive heart transplants using prior experience and their best judgment, says Kilic. However, there are other factors that play into transplant success.

“As a cardiothoracic transplant surgeon, I have to estimate what the trajectory of patients will be before they get a transplant and after they get a transplant,” Kilic explained. “When a donor organ becomes available – those offers are made to clinical teams that are on call and are discussed internally by that institution and by their transplant team.”

In contrast to current methods, Kilic proposes using a data-driven method that factors in as much relevant data as possible, specifically data collected by the National Transplant registry. That data is then fed into a machine learning program that can provide a care team with specific outcome estimates for a potential transplant patient.

“What we're aiming to do is utilize a plethora of data that's available,” Kilic said.

“For example, when a doctor gets a donor offer, rather than just having a subjective discussion with colleagues, they're able to pull up an easily accessible tool that will tell them what is likely to happen to that candidate over time if they reject or accept the donor. So, they're able to balance the consequences of that decision to make a better informed decision,” Kilic explained.

Kilic and his team are also working to adapt an artificial intelligence framework that currently runs the incompatible kidney exchange program in the U.S. to heart transplantation. He is confident this system will be more efficient and better able to handle the dynamic and complex transplant environment where donors and recipients are matched. 

After these tools are developed and passed on to the users, Kilic will work with those end-users to make sure that the tool is working efficiently and effectively. 

Ultimately, Kilic’s goal is to leverage artificial intelligence and machine learning to optimize allocation of scarce heart donor organs, and once allocated, to optimize the decision making of accepting or rejecting these donors for candidates awaiting heart transplantation. He believes this innovative approach can lead to a more efficient process and better patient outcomes.