Cedars-Sinai is working to fill in the gaps in mechanisms, diagnostics, risk assessment and therapeutics of major human disease conditions using artificial intelligence.
The Department of Medicine at Cedars-Sinai created a division to explore AI's applications in healthcare. The division, called Artificial Intelligence in Medicine, was launched March 1 and is led by Sumeet Chugh, MD.
AIM researchers access data from the Cedars-Sinai Health System's clinical data warehouse, using machine learning and deep learning methods to create its own tools and algorithms that have the potential to enhance human disease prevention.
Although the group is in its early stages, it has already produced significant results.
In a study published in the Journal of Nuclear Medicine, AIM researchers employed AI algorithms to identify heart attack risk in patients with already established coronary artery disease.
The AIM division also published research in JAMA Cardiology.
In the study, researchers created an AI tool that can effectively identify and distinguish between two life-threatening heart conditions that are often easy to miss — hypertrophic cardiomyopathy and cardiac amyloidosis.
Dr. Chugh spoke to Becker's Hospital Review about the new division and how his team of researchers is designing clinically relevant questions from the broader Cedars-Sinai Health System that can be ethically vetted, analyzed, validated and implemented using AI.
Q: Why did Cedars-Sinai invest in an entire AI division?
Sumeet Chugh, MD: We're not going to be complete without the ability to utilize this vast storehouse of clinical information that is in our electronic data warehouse unless we have a notice and a core for applied artificial intelligence. So the need is really that we would rather not give our data to somebody else, but rather work on it ourselves, so that we can help our patients in our health system and perhaps possibly others by the discoveries we make.
Q: What is the division focusing on, and how is it leveraging AI currently?
SC: I'm a cardiologist, a specialist in heart rhythm disorders. And so when I was asked to take this on, it's natural that the area that I would focus on to begin with is heart disease.
So, I have recruited three colleagues. Each of us have our groups, our laboratories and staff, focusing mostly on diagnosis and precision medicine, for heart disease.
Our next step is that we are expanding to deep learning in genomics that will extend beyond heart disease to other disease conditions because that's the kind of work that lends itself to cross disciplines.
As a division, we have a very broad mandate, and we're looking forward to expanding to other areas, both in medicine and in surgery.
Q: How do you determine what algorithm to create or what research the AI division needs to do in order to meet patient needs?
SC: So there are two ways in which that can be determined.
As clinicians, we are aware of the gaps in scientific knowledge that exist. And so, there are critical needs out there based on the disease area. For example, for cardiac arrest, there is a critical need to predict who's going to have to make the defibrillator more sustainable in terms of cost. So, as clinicians and as investigators, we are aware of those knowledge gaps.
Another way is that it comes from day-to-day practice. As healthcare providers, as people who read echocardiograms during the course of clinical practice, we will identify deficiencies in our practice. For example, we may miss a diagnosis of a rare form of wall thickness and wish we could go back and change that. So, at least if we can't do that, then we can help patients in the future. So it's a combination of being aware of the knowledge gaps and doing day-to-day clinical practice.
Q: How do you exchange your knowledge and findings from the division?
SC: Every two weeks, we actually do a virtual AI forum where we basically invite outside speakers to share their science and discoveries. Sometimes we have internal speakers, sometimes people present, so we call that the AI forum. And so it's a scientific forum where we share knowledge freely, within the institution and outside the institution.
It is important to involve stakeholders, such as physicians in the early process, so we can get feedback and provide them with regular updates, as well as their own role in the deployment of AI tools.
Q: What are some misconceptions regarding AI in healthcare?
SC: A question that I get a lot from people is, is it going to put physicians out of business if AI takes over?
What I see in the future is not a reduction in physicians, but an increase. I don't think we're going to be letting go of cardiologists because there's AI, we're just going to be hiring cardiologists who know how to deploy AI.