Inside Endeavor Health's AI Echo Adoption: A Conversation with Dr. Ilya Karagodin

Dr. Ilya Karagodin is a general and advanced imaging cardiologist at Endeavor Health in the Chicagoland area, and the Director of AI and Advanced Analytics at Endeavor. His team has integrated Us2.ai's AI-powered echocardiography analysis directly into their native reading environment, driving measurable improvements in efficiency, sonographer experience, and early disease detection. In this interview, he will discuss what it has meant for sonographers, cardiologists, and patients.

 

What was the main motivation that encouraged you and your team to adopt AI Echo into your workflow?

Dr. Karagodin: Yeah, that's an important question. I would say we have the same issue that a lot of echo labs have across the country, which is that we have increasing volumes with insufficient staffing at times. This is a problem that most echo labs are trying to solve. We were motivated to essentially adopt a tool that would help us offload some of our volumes and alleviate some of our sonographer and physician burnout, while helping to improve quality as well.

 

You ended up using the Us2.ai platform, and you integrated that with your Siemens Syngo Dynamics PACS. Could you give us maybe a little bit more insight into how the measurement accuracy and consistency was, and what role that integration played in embedding it in your existing workflows?

Dr. Karagodin: Sure. In terms of integrating with the Siemens Syngo platform, I think that was a key element of the project's success. Being able to work within your native environment, and having all the AI measurements go directly to the native environment via SR mapping, is really key. You don't want to be clicking out into an external platform every time you're reading echoes. You want to be working within your comfortable environment.

That's really what we created here at Endeavor Health, with the help of both Us2.ai as well as the engineers from Siemens Syngo. What we eventually ended up with is having the AI measurements map directly into our viewing platform, which allows us to have a seamless working experience without having to click out in any way, shape, or form.

 

When you're talking about integrating AI into your workflows, data is a big part of it. Did you have any data partnerships prior to this Us2.ai integration that may have helped make this possible?

Dr. Karagodin: Yes, absolutely. Here at Endeavor, we had previously worked with Viz.ai for a number of other use cases. We were able to leverage that relationship, as well as the partnership between Viz.ai and Us2.ai, to get the data pathway approved.

The partnership between Us2.ai and Viz.ai allowed us to seamlessly send the data in an anonymous fashion to Us2.ai and get it back in a way that was acceptable and completely anonymized for our health system. That was a very helpful partnership, and I think it's important to leverage existing partners when trying to employ new technologies.

 

You and your colleagues published an editorial in the Journal of the American Society of Echocardiography last year on user perceptions of AI echo. Could you speak to the key findings of that work, and what guidance would you offer to other institutions looking to drive successful adoption of AI echo among both sonographers and clinicians?

Dr. Karagodin: I'm really glad you asked that, because I think that editorial actually illustrated a couple of important points. Even early on, I think our cardiologists and echo readers were very enthusiastic about the technology. They felt that it was accurate, it was helpful in their decision-making, and most of them would reference the AI when reading the echoes.

On the other hand, sonographers were more skeptical, and there are a number of reasons for that. Obviously, sonographers train for many years to get to where they are. There is certainly perhaps a feeling that maybe their job security is in question or that they're being replaced in some way, shape, or form. That skepticism was definitely apparent from the beginning.

I think the idea of change management is very important. Really talking to and relating to your sonographers is key to empowering them to focus on what they do best, which is image acquisition, and allowing the AI to perform a lot of the manual, laborious tasks that are less appealing and take up more time. That editorial was very helpful in understanding how we can better create this change in our echo lab, and it showed that we should be meeting with our sonographers on a regular basis to understand their perceptions and their feedback.

 

From your perspective, has the adoption of AI echo translated into meaningful time savings or workflow improvements for your team?

Dr. Karagodin: Absolutely, and this is probably the most exciting part of this whole initiative. In recent months, we launched our AI-based workflow where the sonographers essentially acquire minimal manual measurements and allow AI to do the bulk—90 plus percent—of the measurements independently. We compared the AI workflow to our standard manual workflow in terms of time. What we found is that the AI workflow saves, on average, 10 to 12 minutes per study, for a total savings of about 33% per study, which was pretty remarkable.

While 10 or 10.5 minutes may not seem like a lot for one study, when you add that up over thousands of echoes per year, you end up with pretty tremendous time savings. That time can be allocated for other tasks, it can increase throughput through the echo lab, and it can help reduce staffing strain. This was a very promising result that we hope to publish very soon.

 

We've seen concerns from sonographers in the past about how AI will impact their quality of life or their work. From your perspective, what did sonographers think of this technology once you integrated it into the workflow?

Dr. Karagodin: As I mentioned earlier, early on, the sonographers were understandably skeptical of the AI technology. But when they had the chance to actually try out the AI workflow—which, again, entailed taking fewer manual measurements and letting the AI obtain most of them—they had a shift in perspective.

When surveyed, sonographers of all different skill levels stated that they felt they were saving considerable time. They felt that it saved them at least 5 minutes of time, when in fact it was about 10 on average. They felt that their click burden was significantly decreased, and they noted that echo lab efficiency was significantly improved. Perhaps most importantly, all of them voiced a preference to continue using the AI workflow moving forward.

This was very valuable information. It's incredibly important to have buy-in from all members of the echo lab, with sonographers being at the top of that list. To see them have such a favorable response to the AI workflow was really gratifying.

 

Time is one aspect of it, but another aspect of echo is that, especially on the cardiovascular side of things, there are a lot of conditions that are underdiagnosed in clinical practice. One good example of that is cardiac amyloidosis. Given that Us2.ai carries disease detection indications across a range of conditions, do you believe that AI-assisted echo has the potential to meaningfully shift detection rates, and what would that actually look like for patient outcomes?

Dr. Karagodin: I do. I think this is one of the most exciting aspects about AI in the echo lab: we can detect patterns and phenotypes that we previously were unable to detect with the human eye. As has been shown in recent literature, we're able to pick up on disease processes early in their development, even from single echo frames. For example, cardiac amyloid can be picked up from the four-chamber view just based on various features of the echo that, again, are invisible to the human eye.

The ability to detect these disease processes early means that we can get these patients the testing and treatment that they need earlier in their disease process, which really translates to improved patient care over the long run. It can meaningfully move the needle across multiple different disease processes, with amyloid being a perfect example of that.

 

Could you go maybe a little bit deeper on exactly where in the process having this AI detection for amyloid makes a big difference? Earlier detection enables earlier treatment, but what does that do for the patient, and how do they feel about that in your experience?

Dr. Karagodin: Yeah, cardiac amyloidosis is one of those disease processes that can go undetected for many, many years. Often times, these patients can derive the most benefit in the early stages of the disease, before heart failure develops. If you can detect cardiac amyloidosis early on from an echocardiogram—perhaps one that was ordered for a completely different indication—and then you refer that patient to the correct test (such as a cardiac MRI or a nuclear PYP scan) and they get the diagnosis earlier (let's say 3 to 5 years earlier than they would have otherwise), they become eligible for more treatments. They can feel better, and they can live a longer and more full life.

We can really help change the trajectory of their disease earlier on. It has the potential to help a tremendous number of patients, and this kind of early detection is one of the biggest advantages of AI in echocardiography.

 

Before we wrap up, if you could share one message with your colleagues in the cardiology or imaging space, what would it be?

If I had one message to my colleagues—both cardiologists, sonographers, and anyone in the cardiology community—it's to embrace change. Embrace new workflows and new technologies, because change is coming, whether we like it or not.

Ultimately, the individuals who are able to adapt and embrace change are going to be ahead of the pack and ahead of the curve in terms of early diagnosis, echo lab efficiency, and overall throughput. I would encourage everybody to try out AI in their echo lab, see what kind of benefits can be derived, and I think you'll be happy with the results.