AI in Stroke Treatment: Expert Insights from Henry Ford Health

May is American Stroke Month. In this conversation, Aaron Lewandowski, M.D., emergency medicine physician and the emergency medicine stroke representative at Henry Ford West Bloomfield Hospital, and Alex Chebl, M.D., interventional neurologist and director of the Henry Ford Stroke Center and the Division of Vascular Neurology at Henry Ford Health, discuss how artificial intelligence (AI) is revolutionizing stroke care. From accelerating diagnoses and streamlining team communication, to significantly improving patient outcomes, this rapid advancement in AI technology isn’t just supporting doctors — it’s saving lives.



View Transcript
 

00:00:00:27 - 00:00:24:15
Tom Haederle
Welcome to Advancing Health. For stroke victims, speed and survival are closely linked. Quicker diagnosis and treatment can make a huge difference. Coming up in today's podcast, a look at how those two letters we hear more and more about in today's health care - A and I - artificial intelligence, are being applied to protocols for stroke treatment.

00:00:24:18 - 00:00:43:24
Tom Haederle
Hi everyone. I'm Tom Haederle, senior communication specialist with the American Hospital Association and pleased today to get to do one of my favorite parts of this job. And that's highlighting the amazing work that goes on every day among our member hospitals and health systems. And here's a great example: the integration of artificial intelligence into treatment protocols for stroke victims

00:00:43:24 - 00:01:08:11
Tom Haederle
at Detroit-based Henry Ford Health. Joining me from Henry Ford to talk about this are Dr. Aaron Lewandowski, an emergency medicine doctor and the emergency medicine stroke representative at Henry Ford West Bloomfield Hospital, and Doctor Alex Chebl, a vascular and interventional neurologist and director of the Henry Ford Stroke Center and the Division of Vascular Neurology. Doctors, thank you both for joining us on this Advancing Health podcast today.

00:01:08:11 - 00:01:09:08
Tom Haederle
Appreciate you being here.

00:01:09:09 - 00:01:10:07
Aaron Lewandowski, M.D.
Thanks for having us.

00:01:10:09 - 00:01:11:10
Alex Chebl, M.D.
Thank you for having me.

00:01:11:12 - 00:01:20:03
Tom Haederle
Dr. Lewandowski, let's start with you and a basic question: why is speed of diagnosis and treatment so critical when treating victims of a stroke?

00:01:20:05 - 00:01:40:23
Aaron Lewandowski, M.D.
There's a common saying in neurology and stroke care that time is brain. It is estimated that millions of neurons are irreplaceably lost each minute during an ischemic stroke. So the sooner that we are able to diagnose and treat a stroke, the more brain we're able to save and the patients are able to have a easier outcome and a better recovery.

00:01:40:25 - 00:01:45:28
Tom Haederle
And what exactly does AI lend to the process? How has it improved how we're doing this now?

00:01:46:00 - 00:02:12:10
Aaron Lewandowski, M.D.
AI has been used in multiple ways across medicine. In stroke care particularly, we're able to use it in helping with diagnosis of stroke in a timely manner. Our program specifically is called Rapid AI. It is a software program that allows for quicker diagnosis of strokes and also facilitates communication between physicians. Dr. Chebl was actually the physician that brought the idea to our stroke committee, and we've been using it for approximately two years.

00:02:12:12 - 00:02:23:19
Tom Haederle
Does it actually paint - and this is a question for both of you - does it paint a picture of what's going on inside the stroke victim inside the brain actually allow you to see something you couldn't see before. Dr. Chebl?

00:02:23:21 - 00:02:44:01
Alex Chebl, M.D.
It's not so much as paints a picture as gives you an exact picture of what's going on. So the challenge we have in stroke neurology, unlike, say, when a patient comes in with a heart attack, you know, a patient grabs a chest, they're having chest pain. You can do an EKG and a cardiologist emergency physician can know immediately where the problem is.

00:02:44:03 - 00:03:08:12
Alex Chebl, M.D.
The trouble in neurology, is that there are many different types of stroke. Some types of stroke are caused by bleeding into the brain. But the more common type of stroke and why we use AI most commonly is called a ischemic stroke where there's a blockage, and the treatment for those two types of stroke are exactly opposite. One causes the other, and so you have to know what type of stroke you're dealing with.

00:03:08:18 - 00:03:17:15
Alex Chebl, M.D.
And this is why it's more complicated. And knowing what's going on inside the brain with the arteries is critical. And this is where the AI helps us.

00:03:17:17 - 00:03:52:12
Aaron Lewandowski, M.D.
Particularly with ischemic strokes, the issue is trying to figure out what part of the brain has been affected by the stroke and also where the blood clot is. And, is it amenable to intervention? There's medicines such as TMK which we're able to use to try and break down the clot during an ischemic stroke. But particularly where I used it for our purposes is in the use of the thrombectomy procedure, which is where you're able to intervascularly go up into the brain and actually remove the clot that's causing the stroke if it's located in an appropriate and amenable position.

00:03:52:15 - 00:04:15:02
Aaron Lewandowski, M.D.
So the program serves multiple purposes. The AI portion of the program evaluates the CT angiogram and the CT perfusion studies of the patient looking for any asymmetry in blood vessel distribution or perfusion. This is able to allow us to quickly evaluate for signs of what we call a large vessel occlusion. Those are the types of strokes that are most amenable to the thrombectomy procedure.

00:04:15:04 - 00:04:24:03
Tom Haederle
How much time has the use of Rapid AI shaved off of the diagnosis and allowed you to figure out accurately what's happening?

00:04:24:06 - 00:04:51:18
Alex Chebl, M.D.
Approximately 30 minutes. When we look at patients who are candidates for mechanical thrombectomy, that's the procedure where we pull the clots from the brain. We've reduced our door-to-puncture time. That is, from the minute the patient arrives in the emergency department until we actually puncture the artery to get to the brain, we've been able to save about 30 minutes, bringing us down to within the 90 minute ideal window for that treatment.

00:04:51:25 - 00:05:13:01
Alex Chebl, M.D.
But, just as importantly, it's also helped us with our door-to-needle time. So that balloon scan mentioned that you can also give the clot busting medication. That has to be given within 4.5 hours. And so we've now are consistently able to treat patients instead of roughly within an hour presentation. We're now being able to treat almost all patients with 45 minutes.

00:05:13:01 - 00:05:19:16
Alex Chebl, M.D.
And we're approaching 30 minutes from door-to-needle. And every minute is essential in that effort.

00:05:19:18 - 00:05:22:27
Tom Haederle
That's really impressive. What's been the impact on patient outcomes?

00:05:23:04 - 00:05:44:13
Alex Chebl, M.D.
Tremendous patient outcomes. If you look nationally, but also at our sites, you look at the number of patients, proportion of patients who recover to normal or nearly normal has increased. If you look at the number of patients who are discharged to home rather than to rehab, a good measure of whether patients have disability, that has also increased.

00:05:44:15 - 00:05:58:13
Alex Chebl, M.D.
And nationally, the data clearly support, this overwhelmingly so, so that the American Heart Association, for example, keeps shortening the time metric, because the sooner we do it, we're getting better outcomes.

00:05:58:15 - 00:06:17:21
Tom Haederle
Really good news for patients. I'm wondering, given the size of Henry Ford, a big, big system you have. And I imagine that rolling out any new technology or software or changing how things are done, particularly across a scale like that, has got its challenges. Did you run into any kind of bureaucratic obstacles or resistance? We don't know what this thing is . . .

00:06:17:21 - 00:06:21:23
Tom Haederle
Prove it to us. Was it hard to sell, or not really?

00:06:21:26 - 00:06:45:08
Aaron Lewandowski, M.D.
What? Dr. Chebl first brought the idea to us at the West Bloomfield emergency Department, it was certainly interest in, you know, ways that we can improve our stroke care. I would say overall, we didn't really experience any significant barriers to implementing Rapid AI here at Henry Ford. I would say the hurdles that we faced were the standard hurdles you faced with integrating any new piece of software or technology into your preexisting hospital system.

00:06:45:10 - 00:07:23:24
Alex Chebl, M.D.
Yeah, I would second that. You know, there was some trepidation amongst some team members. You know, our implementation of Rapid AI, there's many different ways that you could implement such a program. One could be it just notifies the radiologist, "hey, there's a potential stroke. Take a look." We have gone to the exact or most extreme or the deepest implementation, meaning all members of the team are notified when we have a stroke, and this has minimized the number of phone calls we have to make to get the patient ready, to get the OR team ready, etc. and when you have that many people learning something new there can be some trepidation.

00:07:23:24 - 00:07:44:12
Alex Chebl, M.D.
And the biggest fear really was, why do I have to have another app? And this is just going to increase my workload, right? I'm going to be bothered all the time with these unnecessary things. And in fact, it's the exact opposite. Most people got used to it. They could not believe that they were living without it. It's made their lives better.

00:07:44:12 - 00:07:49:11
Alex Chebl, M.D.
Not just the patients lives better. It made all of our lives better because it's simplified the communication.

00:07:49:14 - 00:08:21:26
Aaron Lewandowski, M.D.
And I would certainly second that. From an emergency medicine perspective, a lot of our job on a day to day basis is discussing phone calls with consultants and trying to communicate with other team members. So being able to have that initial phone call with the stroke neurologist to discuss the initial plan of care, but then everything else being in the, HIPAA secure chat with rapid AI has certainly allowed for our communication to be much more effective and much more quicker so that everyone can see in real time what's going on, what's the plan?

00:08:21:26 - 00:08:23:14
Aaron Lewandowski, M.D.
What are we doing for the patient?

00:08:23:16 - 00:08:44:22
Tom Haederle
Yeah. You hear that so often about applications of AI and in almost any capacity, ambient listening or anything else. People are delighted. It's a time saver and a work saver. And you've seen that with the with the implementation of, Rapid AI at Henry Ford. Any thoughts you would share about another system or hospital that is considering going around and maybe integrating it for the first time?

00:08:44:25 - 00:08:50:24
Tom Haederle
What would you say in terms of it's utility, in terms of its ease of use, that kind of thing?

00:08:50:26 - 00:09:17:29
Alex Chebl, M.D.
Well, I mean, I think there's two aspects. One is you've got to lay the groundwork for this. You need a stroke champion, champions. Certainly someone from emergency department is critical. You need someone on the neurology side. And they need to then sell this to everyone. Once you've laid the groundwork and you've got buy-in from everyone

00:09:18:01 - 00:09:41:20
Alex Chebl, M.D.
the actual implementation isn't that difficult. Securing IT, and the firewalls, etc.. The company helped set up. They also have individuals who can come and help train users. How to use it, how to adjust the settings, etc.. So we found that it was pretty straightforward to initiate the Rapid AI in our system.

00:09:41:26 - 00:10:00:29
Alex Chebl, M.D.
And one way to do it, I guess, would be my suggestion would be don't start too big. You know, maybe start if you have a large system like we have, you know, start locally, 1 or 2 smaller hospitals. Don't include every single team member. Get the bugs worked out of the system and then expand.

00:10:01:01 - 00:10:21:13
Aaron Lewandowski, M.D.
And definitely when you're trying to, you know, sell the idea to administration or other departments, certainly focusing on the benefits to patient care, like quicker diagnosis and also the benefits to the team members, such as more effective communication. I think is a really good way to show the positive benefits that can come from this.

00:10:21:16 - 00:10:47:18
Alex Chebl, M.D.
You know, obviously we do everything focused on the patient. We want the best patient outcomes, but we can't deliver good health care without paying for everything that's required to do so. So the money does play a role. And I think this is where it's important for an administrator to understand is that the better the patient does, the shorter length of stay, the less money is spent on that patient.

00:10:47:22 - 00:11:02:22
Alex Chebl, M.D.
And therefore a health system can keep more of that money for the other services that they need. And I think that's very important. I mean, after all, this is why we were able to convince CMS to pay for these very complex treatments is because overall it ends up saving money.

00:11:02:24 - 00:11:09:23
Tom Haederle
It's a great point, thank you. As we wrap up, any final thoughts? Anything we haven't talked about that you'd like to say about Rapid AI?

00:11:09:26 - 00:11:35:20
Alex Chebl, M.D.
You know, these systems now? Although they're mostly started in stroke, there are many competitors, Rapid AI as well, but they have other modules. And so these systems can be used for other disease states, pulmonary embolism, the identification of intracranial hemorrhage, cerebral aneurysms. And so there are many opportunities for multiple different departments to collaborate. And that can also help with the financial aspects of this.

00:11:35:21 - 00:11:46:14
Alex Chebl, M.D.
You know, the more users you have on board, it tends to be, you know, cheaper than just having each individual division having their own systems working independently.

00:11:46:16 - 00:11:51:12
Tom Haederle
That's a great point, thank you. Thank you for bringing that up. Dr, Lewandowski, any final thoughts?

00:11:51:15 - 00:12:09:23
Aaron Lewandowski, M.D.
I've certainly enjoyed the implementation of Rapid AI. It makes my job simpler. It provides better patient care. You know, I don't think that AI will ever replace physician assessment and judgment, but it's very impressive what a powerful tool it can be when used appropriately, to improve the care that we provide to our patients.

:12:09:26 - 00:12:21:22
Tom Haederle
Absolutely. Thank you both so much for your time today and this great discussion. And I hope it reaches a lot of ears and get some people thinking about just how powerful this tool is. So again, appreciate your time. Thank you for being on Advancing Health.

00:12:21:25 - 00:12:22:15
Aaron Lewandowski, M.D.
Thank you very much.

00:12:22:20 - 00:12:25:05
Alex Chebl, M.D.
Thank you. Have a wonderful day.

00:12:25:07 - 00:12:33:18
Tom Haederle
Thanks for listening to Advancing Health. Please subscribe and rate us five stars on Apple Podcasts, Spotify, or wherever you get your podcasts.