Artificial intelligence tools have opened new pathways for physicians and researchers to deliver patient care and further medical discovery. Dr. Mark Cohen, dean of the Carle Illinois College of Medicine at the University of Illinois Urbana-Champaign, talked with News Bureau biomedical sciences editor Liz Ahlberg Touchstone about the risks and rewards of using AI tools in health care and medical discovery.
How can AI help physicians care for patients?
There are several ways that AI can benefit physicians in terms of patient care. AI as a tool can provide a lot of information to a physician in real time, which saves them from having to look up a variety of different sources, so it can be a time saver in the clinic. It also can organize information in a way that’s much more presentable and easier to digest, so there are a lot of areas around documentation in the clinic where AI tools could be very helpful. For example, if you had an AI tool that was listening to a conversation between a provider and a patient, it could summarize that conversation into a medical note that could then be put into the record after review.
Moving beyond that, AI tools help with diagnosis in other areas of medicine. One example is in imaging, where a provider can review a lung X-ray, and an AI tool can help them more accurately identify lesions or abnormalities. As these tools continue to develop, I could take an image of a patient’s X-ray or CT scan and convert it into a three-dimensional object. I can look at it in front of me, but someone 3000 miles away who might be the world expert on that disease can immediately see that same object in real time, and we can have a conversation about how to best care for my patient. This can be transformative for rural or underserved areas: being able to bring expertise right to the bedside, to the local emergency department, to where it’s needed at a time when it can really save lives. We’re starting to think about that care delivery and how these tools can really help us address an issue like access to specialty care, which is a real problem both in rural Illinois and around the world.
Does using AI risk losing the human connection between doctors and patients?
We have to think about AI as a tool, not a care provider. AI is not meant to replace physicians. It’s a tool to help physicians and other healthcare providers work more effectively. So if you think of it as a tool, there’s always going to be the human factor in healthcare that’s so necessary for patient trust, for developing a relationship for that treatment process. As AI is being implemented, it can gather information, it can help with organizing lots of information into usable options for physicians, but it’s not going to replace the humanity or human factor of medicine.
What are the risks of relying on AI for medical information? Is it trustworthy?
That’s been a big question. Obviously, protected health information is very important to maintain data security, to maintain that level of trust. As we look at how AI tools are being utilized in the clinical environment, I think it’s really imperative to think about the data safety piece to make sure that protected health information stays protected.
We also need to consider reliability. When AI searches a database, the answer it gives is only as good as the data that was in the database. As long as that data is reliable, you can trust it as a good tool. So the question then becomes, how do we make sure that what’s being searched is reliable data that’s been verified and validated, and that what we’re getting out is something that can really help providers make better decisions?
Similarly, for patients asking an AI chatbot for medical information or advice, one of the challenges with asking a broader engine like chat GPT is that it looks at everything that's out there on the internet. While there is some good information out there, there are also many things that are not so reliable. How the AI pulls that information and determines the quality and reliability of that information is really important.
How could AI help with medical discovery, like drug or device development?
It’s really exciting to think about how data can be used more effectively. We already have data from your electronic medical record, data from different tests and results that you've had over the years, data from genetics. What if AI tools could take all those data sets and make a digital version of you that it can then evaluate? When a new drug is considered, it can ask, is that the right drug for your particular condition, and based on your genetics, your history, is it going to work well? It can run thousands of probabilities and risks all in seconds to say this drug might be a good fit, or suggest what the next best option could be.
Another area where AI can be useful is drug discovery. AI can analyze millions of different types of chemical formulas and create new molecules with medicinal properties. Similarly with devices, thousands of variants can be prototyped virtually to optimize parameters and think about how to apply new devices in patients. AI tools can really help accelerate some of that evaluation process, as well as defining potential new options or pathways for discovery.
This past year, we at CI Med created a new global consortium of innovation and engineering in medicine that brings together medical schools, engineering schools, companies, government agencies and regulatory agencies to think about how as a large, diverse, interdisciplinary group, we can solve some big data problems together. One of those is creating a global de-identified data warehouse that could be an amazing tool. What if AI solutions could look at not just a thousand patients or a million patients, but 500 million patients around the world and solve some real problems around rare diseases, cancer, or some chronic diseases that we are only scratching the surface of with limited data? With new tools like quantum and other ways to make sure that data is safe and encrypted, such an endeavor could take us to the next level of how we think about healthcare solutions and how we think about collaborations across the world.
How is CI Med using AI to train future physician-innovators?
We have an imperative as a medical school to think about the next generation, and technology is only going to become more and more immersed in how we think about our future in medicine. If we’re not training that next generation to use those technologies more effectively, then we are doing a disservice to the future.
CI Med is the world’s first engineering-based medical school. We were founded on the principle that this intersection of technology and engineering is an important part of how to teach medical students how to become future physician-innovators who can lead interdisciplinary teams to think about how to solve bigger problems in health care. Our focus has been really proactive to think about better educating medical students to use tools like AI in all of these different domains.