This Week Health

Today: Looking at AI in Imaging with Lee Milligan, MD

Dr. Lee Milligan sits down with me to discuss AI in Imagining. Hope you enjoy.

Transcript

Today in Health IT, we're going to take a look at imaging and specifically AI and imaging, how it's progressing and where it's going. My name is Bill Russell. I'm a former CIO for 16 Hospital System and creator of This Week Health, a set of channels and events dedicated to leveraging the power of community to propel healthcare forward.

We want to thank our show sponsors who are investing in developing the next generation of health leaders. SureTest, Artisite, Parlance, Certify Health, Notable and ServiceNow. Check them out at thisweekhealth. com slash today. I just want to thank the community again. We set a goal to raise 50, 000 for Alex's lemonade stand.

We're up over 55, 000 for the year. We want to plow through that number. If you want to join us, hit our website, top right hand column. You can click on a logo for the lemonade stand. You can give there today. We believe in the generosity of our community and we thank you in advance. All right. Today, I'm not alone.

I am hanging out here with Dr. Lee Milligan. Lee, thanks for joining me. Thanks so much for having me, Bill. This is only the second Today Show that I actually had a guest on. , but, , you know, I was asked today about AI and imaging and where it's progressed to and whatnot by a CIO. And I just sort of laughed because it's someone who knows you.

And I was like, call Lee. He knows a ton more than me. And I thought, you know what? You do know a ton more than me on this subject. So let's, let's talk about this. A lot going on in AI, a lot going on in imaging, and it's, imaging is one of those areas where it has matured. Like, it went through the early stages a long time ago, and it's starting to mature.

Talk, talk a little bit about what, what we're seeing in this

space. Yeah, I think if you think about all of the different potential spaces for AI to impact medicine, I think imaging is arguably the most susceptible to this, and it's, , it's made some doctors actually quite worried about Essentially being replaced by AI.

, if you think about it, you've got, with images, you have large patterns that directly correlate to a discrete field, an ICD 10 code. And you have millions and millions and millions of these that you can turn over and over and over. And so, I think it's not surprising that there's something like 400 companies.

that are now developing, , diagnostic AI in the imaging, in the medical imaging space. And I would, there's a lot to talk about here, but, , as I've been on my journey with my former company, , as I looked at the opportunities here, I was really trying to understand where it could be the most impactful, while not adding significant additional cost.

And we can talk about that a little bit, because I think those are good business, , discussions to have.

So, do you find... Different things in the administrative side, different things in the radiology, different things in the cardiology

side? Yeah, each of them have kind of a different goal as it relates to the diagnostics of it.

But let's just take x ray, for example. Right. So x ray is, you know, old technology, been around for 150, 50 years, something like that. Single image. Single image, two dimensional image. And, you know, arguably it's one of the worst things that a radiologist would want to read. Right? Because today, they get paid a heck of a lot more to read an CT or a dex or something else than an X ray.

And, because it's two dimensional and it's old technology, arguably there's somewhat more risk associated with an X ray than there is with a three dimensional image such as an MR or a CT. So, the, the reward is low, the risk is high, and therefore folks, you know, are probably not going to be too inclined to want to read these.

But some of the AI that's developed in this space has really been remarkable. And, , there are some companies who have been able to, , create a framework where from a sensitivity and specificity scenario, it's equal to or above the average of high quality humans, radiologists, reading those images.

But, but you mentioned the, the concern early on of, of our, the, the people doing the reads being worried about being displaced.

But that's, that's not really the design construct that we're going for, is it? We're just looking to augment them and have them read those things. Or are we, are we looking at essentially automated reads? I think there's a

lot of discussion around not replacing the radiologists, intended to calm radiologists.

I think the reality is that this technology is going to, across the board, eventually meet and then exceed the, the accuracy of a human being looking at the same images. And so I think that the different question that should be asked is how do we continue to, , Maintain the, the human being in the process as kind of a check and balance while the AI does the bulk of the actual underlying work.

I think that's the real challenge. Yeah, so

this is a short show, so I'm trying to figure out what the, if I'm standing up an imaging center of the future, I'm a health system CIO and I'm thinking, hey we're gonna do things a little different, we're gonna, we're gonna get out there, and they want to take a look at ways that it can be done more efficiently.

More accurately provide better care for the community. What are some things they should be looking at and how How would you approach

that? Yeah, I mean, I would look at the entire business model everything from the Reception of orders coming in right now. Most imaging centers have three different pathways handwritten copies fax , and EHR integrations.

I would look to maximize that entire experience with EHR integrations. , second, I would look at automations throughout the contact center, so that you could automate the process the patient has to go through while they're actually scheduling their appointment. And then lastly, when it comes to AI specific to diagnostic medical imaging, I would look at focusing on operational efficiencies.

So going back to the case of x ray, and in our scenario, we were able to decrease our turnaround times by about 36 hours. And that was meaningfully impactful, materially impactful to the business, , the business model.

36 hours. That's, that's a significant amount of time. It is.

It is. And when your cues, you know, as the business grows and as the cues, , of unread images grows, , you have to figure out ways to, , improve the efficiency associated with those reads.

So, I want to talk a little bit about, , cardiology and MRIs, CTs, just the more complex scans. Are, is, is AI making inroads in that

area as well? Yeah, yeah, huge. So, you know, one of the things that cardiologists look at when they're trying to determine whether there's disease in a patient is, number one, is their plaque in the artery itself.

And then they have to quantify that plaque. Is it, you know, is it a third of the, of the size of the lumen? Is it half the lumen? Is it 90 percent of the lumen? And, , so that's one aspect of it. The second is the character of the plaque itself. In the old days, they used to think that all plaques were created equal.

It turns out some plaques are more susceptible to rupture, and when they rupture, that's when platelets adhere and form a clot, and that's when a heart attack actually happens. But the other big thing that they look at is the, , what's called the FFR, the fractional flow reserve, and that previously could only be measured by putting somebody in the hospital.

Running a catheter in their groin, up to their heart, releasing dye, taking pictures, and then on top of that, , snaking a wire past the plaque itself to measure the pressure inside the artery past that plaque. Today, you can do that entire thing with imaging. Which tells me that, you know, really in the near term future, cath lab is if you're gonna have an actual procedure performed to open something up and to keep it open, versus diagnose what the actual problem is.

It's pretty cool.

The other use case I found interesting was more of a population health use case where, you know, you get these models in place and you can take images for the last 10 15 years and process those images and say that was a misread, we missed, you know, this or we're going to do outreach to these types of patients and those kinds of things and you can process a fair amount of x rays, obviously, very quickly.

, going that path. Oh,

yeah. I mean, there is one model of doing this where you have the, the, the human board certified radiologist read on the front end and you simply run QA on the back end and then you, you identify specific cases and you pull those and have your chief medical officer or who have you, , take a look at those and review those.

Have you, have you heard of, , the approach of going back 10 years and looking at your population and seeing if there's... Anybody can reach out.

I don't, I don't see the value in that, honestly. It's too far back for a couple of reasons. One, the imaging technology has changed drastically. And two, the relative value of a finding from 10 years ago is marginally quite small.

Interesting. So, imaging. Imaging has changed pretty dramatically over the last 10 years. Where will the change come in the next 10 years? Is it going to be in the imaging equipment, or is it going to be more in the... The, , the systems that we're going to be using to do the reads.

I think both. I think that the changes that have happened have been both good and bad, frankly.

So, you know, the good obviously is the enhanced technology that gives sharper images, better quality images, and hopefully more accurate reads ultimately, and more efficient around kind of how we process images. , I think the downside is that it's become more transactional. So in the old days, you know, I'd be in the ER, and I'd have some really crazy abdominal pain scenario with a patient.

And I would, based on whatever's going on, I might be worried about something very specific. Like a splenic artery, , embolism, or something like that. Something really rare. And I could just walk down the hall and say, Hey Mac, I got a really weird case. Here's the one thing I'm worried about. Can you take a look at that and just give me a sense whether that's a possibility or not?

And then together we would look at it. Review it. And I would arguably be more comfortable discharging the patient if I knew that the radiologist reading that was A, good, and B, understood my concern and directly addressed my concern. Then I could actually be more efficient about how I, , disposition that patient.

That piece, unfortunately, has gone the way the dodo bird, as it relates to, , to imaging. In terms of where we're going to go, , technologically, I think we're on the very cusp of huge things in this space. And I think the smart health systems and the smart imaging companies are going to be, , thoughtful about how they position the human being while they're surrounded by

AI.

That's interesting. Lee, always a pleasure to catch up with you. Appreciate it. Love it, Bill. Good talking to you. All right. That's all for today. Don't forget, share this podcast with a friend. Keep the conversation going. We want to thank our channel sponsors one more time. We thank them for investing in the next generation of health leaders.

SureTest, Artisite, Parlance, Certify Health, Notable, and 📍 ServiceNow. Check them out at ThisWeekHealth. com slash today. Thanks for listening. That's all for now.

Want to tune in on your favorite listening platform? Don't forget to subscribe!

Thank You to Our Show Sponsors

Our Shows

Keynote - This Week HealthSolution Showcase This Week Health
Newsday - This Week HealthToday in Health IT - This Week Health

Related Content

1 2 3 240
Transform Healthcare - One Connection at a Time

© Copyright 2023 Health Lyrics All rights reserved