Michael Denning is Diagnostic Imaging Manager at a major US cardiology practice in the Midwest (Wichita, Kansas), with 25 years of experience in echocardiography. In this interview, he shares his first-hand experience adopting Us2.ai in a busy real-world cardiology setting, from the initial motivation and deployment through to measurement accuracy, time savings, sonographer buy-in, and his advice for other practices considering AI echo.
So, you recently adopted an AI echocardiography solution in your practice. Can you tell us a little bit about what the solution was and maybe what was the motivation that prompted the consideration of AI echo?
Michael Denning: The product is Us2.ai. It is a solution that simply takes images acquired, runs them through the AI, and performs all the measurements automatically. The reason we looked into this solution is, as for many practices, staffing challenges in ultrasound, specifically in echo, can be very difficult. Backlogs were becoming a real problem, patient access to their echoes, cardiac clearances, and so on. As we trialled it, we really saw the benefit of being able to save time by not measuring while scanning, and it offered a solution to help with some of the backlog.
And echocardiography requires a lot of work on the part of the sonographer, doesn't it? I mean, there's a lot of measurements that have to be taken.
Michael Denning: It does! There are many measurements that we perform. Traditionally, you scan for two or three minutes and you perform a set of measurements, then you scan another few minutes and perform another set of measurements. With Us2.ai, we are able to scan from start to finish without a single measurement. It sends to our PACS system, runs through the AI, and when we come out of the room it is completely measured. We give our preliminary findings and our job is complete at that point.
Sounds like that should be a lot more efficient! Now how did Us2.ai perform in terms of measurement accuracy and consistency after you implemented it?
Michael Denning: After the introductory period: getting used to what it looks for to measure and slight tweaking on scanning techniques — I have spoken with many of our techs and cardiologists, and just from an internal perspective, everybody is in agreement that it gets it right 85 to 90% of the time. Nothing is ever 100%, but it is actually quite an impressive tool.
Reporting backlog and turnaround times are persistent operational challenges at Echo labs all over the country. Has the adoption of AI echo translated into meaningful time savings or workflow improvements at your site?
Michael Denning: Absolutely, drastic. Time savings from not measuring saves, on average, 6.5 minutes depending on how elaborate your standards are, it could honestly be anywhere from 5 to 10 minutes. But not only does it save measuring time, it also allows for non-traditional workflows of scanning, turning rooms around, scanning another patient, and then coming back to finishing preliminary reports at a later time.
Is Us2.ai actually enabling your (echo) techs to do their jobs with fewer clicks, holding the probes in their hands less?
Michael Denning: I haven't actually counted the number of clicks that it saves, but it saves tremendous amounts of clicks. As you know, holding the transducer up against the patient, that's probably 6 and a half minutes that it saves truly with patient contact with transducer in hand. It saves on average 6 and a half minutes!
And that's a big issue, the whole how long sonographers are holding that transducer, that's a big issue in terms of ergonomics that can lead to muscle-skeletal issues right?
Michael Denning: Absolutely, that is what leads to the repetitive motion injuries that all sonographers will face at some point in their career. (With AI echo), it absolutely saves time scanning per patient. Over the days, the weeks, the months, the years, it will save hours of scanning time. That is a real and meaningful benefit for sonographers' long-term health and wellbeing.
You mentioned measurements a bit earlier. Is producing clinically better measurements that are less variable (with AI echo a benefit)? Because another issue in ultrasound in general is operator variability, how much of an issue is that?
Michael Denning: Yeah, so (AI echo) has helped to curb operator variability, depending on (echo) tech experience, skill set, it has definitely helped with some of that. The (echo) tech is ultimately responsible for the preliminary findings and measurements that they provide to the physician. So if the AI does get it wrong, we can still view the measurements and adjust it correctly. But again, 85-90% of the time, I'll speak from my personal perspective, it gets it right on the money!
So could you talk a little bit about the actual deployment of the AI echo at your facility? How long did it take? Was there initial resistance between the (echo) techs and the cardiologists about what this new technology was going to do?
Michael Denning: The deployment itself — turning it on — is immediate. Once the connection is made with the help of all the IT people, and they turn it on, it just happens automatically. You scan, you send a study as normal, and when you are looking at it in your PACS system, the measurements are there. That deployment is easy.
Where more thought needs to be had during implementation is (echo) tech buy-in. New technology is always scary. For many places this will result in significant workflow changes. Many (echo) techs have measured strictly on the ultrasound system itself their entire career, so the thought of changing that entire workflow is scary. The thought of AI taking over jobs, all those things are scary.
I think it is very important to discuss with techs and point out the benefits, and to get the mindset that it is a tool for us to use. It is simply something to make us better at our job. We are still scanning, we are still doing the things that AI cannot do. It is simply making our job easier by doing all these measurements. We are still overseeing it, but it is helping us do our job. There is still a human in there, that is an important emphasis.
How long did it take for you personally to become comfortable with the outputs you are seeing from the Us2.ai software?
Michael Denning: Personally, honestly, a week into it and I was pretty comfortable. If you fully embrace it with a positive attitude and outlook and know that it's not going to be 100%, but it's going to hit that 85-90%. It does not that that long to adjust honestly.
I understand that you are using Studycast at your site, can you talk a little bit about what Studycast does and then how Us2.ai integrates with that and how it may have aided the adoption of AI at your institution?
Michael Denning: That is actually what made Us2.ai such an effective tool – it integrates seamlessly with Studycast. The images go from our ultrasound system to StudyCast. StudyCast sends it to Us2.ai. Us2.ai does the processing, measures all of the measurements that it can, measures strain at a high rate, and then sends that information back to StudyCast and populates all those measurements. That whole process takes roughly two minutes. It is very quick. As an (echo) tech, we really don't have to do anything different. We send the images to our PACS system like we always would. This time, when we pull it up, the echo measurements are there waiting for us.
Thanks Michael! Do you have any final advice for practices that are looking to adopt AI echo?
Michael Denning: I cannot stress enough (echo) tech buy-in. It is the techs that it affects the most. It is scary, it is new, it is something we have not seen before. I think promoting it as a positive tool that helps us is going to be paramount, easing that tension of being scared of change.
It also really needs to have a champion in the organisation for it to work. I would probably start with one person, get them very comfortable. Your lead tech would be ideal, somebody that is very open to change, and get them to implement it in a positive way. It is probably the most impressive tool I have seen in my career!