Can Artificial Intelligence remove unintended bias from health care? Clinicians optimistic, but wary

May 11, 2022
Jihad Obeid speaks as two other colleagues listen during a discussion on AI in health care
Dr. Jihad Obeid, center, speaks about the pros and cons of AI in health care, as Drs. Paul Heider (left) and Christopher Metts listen. Photos by Sarah Pack

During one of the many live collaboration panels of MUSC’s 2022 Innovation Week, an interesting discussion ensued, mirroring a common debate in health care and that is: How does artificial intelligence (AI) fit in?

Last week, as several clinicians and key members of the Clemson-MUSC AI Hub – which was formed in 2021 – were on hand at the Gazes Cardiac Research Institute, it became quickly evident that AI is gaining traction throughout the world of heath care. But equally evident was the fact that there’s still some skepticism from the mainstream when it comes to the best ways to use it.

For congenital cardiologist G. Hamilton Baker, M.D., associate professor of pediatrics, AI remains a tremendous untapped resource. 

“AI is such a blanket term,” he said in an interview right after the formation of the Clemson-MUSC AI Hub last year. “We’re leveraging data science and wrangling those giant databases with appropriately applied machine learning methods.”

Baker has been utilizing AI in his work for several years now, working on a number of different AI+Biomedical projects ranging from congenital heart disease to diabetic eye disease. 

“I feel very strongly about education on AI. The goal is to teach clinicians how to understand and utilize AI. We aren’t asking people to learn how to code, we simply want them to learn how AI can work for them,” Baker said. 

At the Gazes, the topic quickly centered on AI and bias. Some clinicians believe the most elegant aspect of AI is that it removes unintended biases by letting the computers – which are inherently without bias because they’re metal and silicone – do the data crunching and leaving the treatment to the physicians. 

Hamilton Baker looks at his computer monitor as he hosts a discussion on AI in health care 
When asked the question, "Would you use an AI Health care solution you couldn’t explain?" Dr. Baker, who hosted the summit, said, "I guess it depends on whether we're talking about a skin cream or brain surgery."

“When two clinicians might disagree on something, AI can help uncover unknown biases and dispel others,” said MUSC Public Health Sciences assistant professor Paul Heider, Ph.D. “AI just looks at the data and makes decisions that are based on that alone.”

However, others argued that those AI programs were written by humans, and those inadvertent biases almost certainly were sprinkled in. 

“Trustworthiness is a key word that we need to be focusing on here,” said Brian Dean, Ph.D., chairman of the Division of Computer Science at Clemson University. “Because the AI system is becoming less of a smart sensor that provides input to the medical decision-making process and more of a teammate. So we have to be super careful because, after all, AI was trained based on human expert opinion, which is biased.”

Dean agreed that AI is an extremely valuable tool for the medical field, cautioning all to simply be judicious with its use.

Jihad Obeid, M.D., co-director of the Biomedical Informatics Center at MUSC, agreed. “If you use it as a decision aid, rather than a decision-maker,” he said, “AI can be a real asset.”

Regardless of the differences of opinion in the room, panel members agreed that AI has unlimited potential for researchers and clinicians alike.

“When it comes to AI in health care, it’s so tempting to talk about the hype, all the big stuff it can do,” Baker said. “But the truth of the matter is there are plenty of easy, smart projects where AI could really make a significant difference, and we just need more people on board.”

According to MUSC provost Lisa K. Saladin, PT, Ph.D., MUSC is already using AI to develop techniques that can help to diagnose and treat a range of ills, including cancer, Alzheimer’s disease, substance abuse, child abuse, epilepsy, aphasia, inflammatory skin conditions and cardiac issues.

Baker said that clinicians who are interested in implementing AI into their research or practice should look into the AI Hub, as it offers a host of resources, including funding for AI. During this year’s Innovation Week, the Clemson-MUSC AI Hub gave out $100,000 worth of grants to five worthy projects.

“We want people to know about this,” he said. “I know there are lots of people out there who could really use our help. We want to accelerate the adoption of AI for those who are interested."