BackTable / VI / Article
AI for Aortic Injury: Emergent vs Elective Care
Caleb Solivio • Updated Jun 25, 2023 • 35 hits
We use artificial intelligence (AI) in our everyday life without even knowing it; when we Google the nearest restaurant, ask Siri about the weather today, or as we scroll through social media on our phones. But what if AI could be applied to life-threatening conditions? Dr. Ben Starnes, vascular surgeon at the University of Washington (UW), weighs in on this topic. He suggests that the incorporation of AI into the management of aortic injuries would be revolutionary. AI could help to modernize a system that is clunky and outdated into one that is more efficient and proactive rather than reactive. This article features excerpts from the BackTable Vascular & Interventional Podcast. We’ve provided the highlight reel here, and you can listen to the full podcast below.
The Backtable Brief
• The current approach to managing aortic emergencies is cumbersome, especially across hospital systems. It requires the careful coordination of physicians who diagnose aortic emergencies, physicians who treat aortic emergencies, multidisciplinary teams involved in patient management, and other intermediaries who ensure smooth patient transfer and access to patient medical records.
• In emergencies, Artificial Intelligence (AI) programs have life-saving potential by dramatically reducing the workup time for aortic injury. AI healthcare platforms such as Viz.ai can aid in accurately detecting aortic injuries, rapidly disseminating important information across various health care professionals, and seamlessly coalesce imaging from various healthcare systems into one place for viewing on mobile devices.
• It is not uncommon for a slow-growing or small aortic aneurysm to go untreated, especially in the face of more urgent health concerns. For this reason, AI can be extremely helpful in sorting through large volumes of medical documentation to generate lists of patients with untreated aortic injuries to be followed up with and subsequently treated at appropriate times.
Table of Contents
(1) Aortic Emergencies: Diagnosis to Treatment
(2) AI Recognition of Emergent Aortic Injuries
(3) AI for Elective Aortic Injuries
Aortic Emergencies: Diagnosis to Treatment
Upon diagnosis of an aortic issue at UW, such as a rupture or a dissection, physicians typically contact a centralized service known as the Transfer Center, which then connects with the on-call vascular surgeon. The surgeon reviews the patient's images, gives advice on how to manage the patient, and prepares for the patient's arrival. Uniquely, Dr. Starnes's practice covers aortic emergencies for five states, but this increased coverage leads to difficulty in accessing patient images, especially if patients are transferred from hospitals out of the UW hospital system. Thus, addressing aortic emergencies is a complex, cumbersome process that involves asynchronous communication across multiple systems to ensure efficient emergency awareness, access to imaging, and preparation for patient transfer.
[Dr. Sabeen Dhand]
Let's talk about an aortic emergency, whether it's a rupture or a dissection. When the diagnosis is made, usually I'm guessing by the radiologist reading the scan in the ER, what's the next step in the workflow at UW before any kind of platform or anything? What would happen? Would they call the ER, would they call you?
[Dr. Ben Starnes]
The current workflow is pretty cumbersome and outdated, in my opinion. Normally, we'll get a call from a physician in an emergency room somewhere in the state where he says, "I have a 74 year old patient who presented with abdominal pain. We got a CT scan, and we found an eight centimeter ruptured AAA." and then basically, they will call the Transfer Center, which is kind of a centralized service, who gets in touch with the vascular surgeon on call. We will then have to find our way to a computer somewhere, typically a desktop but not on your iPhone, to look at the images.
We then look at the images once we've gotten on the computer, and we accept the patient and transfer. Sometimes we'll give the referring provider some tips on how to manage the patient, whether it's blood pressure regimens, that sort of thing, and then the patient starts on their way, either by ambulance or by rotary aircraft, and then we'll prepare for the patient to arrive. We'll have the images to look at and prepare for the case by grabbing the appropriate graphs, that sort of thing.
[Dr. Sabeen Dhand]
Are these transfers within just the four hospitals that you're working on, or are all the community practices around Seattle or in that area? How big is your catch zone?
[Dr. Ben Starnes]
It's interesting because we work in a very unique place. We're the only level one trauma center for five American states, Washington, Alaska, Wyoming, Montana, Idaho. That's 27% of the landmass of the United States, but only 15 million people, so we get calls from Montana, we'll get calls from Anchorage, Alaska, where they have very few vascular surgeons available. I think they only have one or two in Anchorage. We'll get calls from all over the state, all over the five-state region. When it's within our own system it's pretty streamlined. All of the aortic emergencies will come through Harborview mostly. We'll take those patients directly to the operating room and fix their problems.
[Dr. Sabeen Dhand]
One thing you mentioned, seeing the images. From my radiologist's standpoint, are all the images on the same PACS system? How do you have access to all that five states' wide? That's hard.
[Dr. Ben Starnes]
That becomes an issue as well. It's taken about 10 years for us. Every time someone calls, if they're not part of our network, they are by the end of the evening, because the transfer center will be obligated to get them to sign in to our PACS system and have our PACS systems be able to communicate seamlessly. Sometimes it boils down to the fact that we can't get the images across and I'll have that provider, whether it's an ER physician or another surgeon, take snapshots of the CT scan with their iPhone and just text them to me. I know that's old school. It's not HIPAA compliant at all.
[Dr. Sabeen Dhand]
It’s cumbersome. That happens. It's very hard. I know from a technical standpoint from PACS that it's not that easy, or it takes some time to get those images to you. It's a very important fact there about images because we'll round back to that. What about contacting? We're going analog, we're calling you on a cell phone, transfer center. Who else is being notified other than you? Are the ICU hospitalists being notified or are you —
[Dr. Ben Starnes]
Not at all. They're usually surprised by the fact that we have a patient that's post-op from a ruptured AAA.
[Dr. Sabeen Dhand]
Exactly. Who's calling the OR, is it you personally calling the OR/surgery too?
[Dr. Ben Starnes]
Typically me, a resident, or a fellow is calling the front desk of the operating room to let them know. It's multiple phone calls. It's not seamlessly integrated at all.
[Dr. Sabeen Dhand]
Exactly. This is very similar to what we were dealing with, with stroke. Even right now with our aortic emergencies at my hospital, again, like you were saying, ICU doesn't even know until the case is done and we're calling them. It's a lot of these analog connections now. What about reps? Are reps frequently supporting your cases or you have all the grafts available at your hospital and you do all of that on your own?
[Dr. Ben Starnes]
We're lucky because we treat. Last year we did more than 350 aortic cases across-
[Dr. Sabeen Dhand]
Wow, that's amazing.
[Dr. Ben Starnes]
-our four-hospital system. Most of those aortic cases are focused in two hospitals, Harborview and UW Montlake. It depends on the physician. I typically don't ask for reps to be in the room. We have over 300 aortic stent grafts on our shelves so we can choose the right graft for the right anatomy, but we're very comfortable in using the grafts. Some of our newer faculty may like to have the reps in the room to help them get through the case successfully. It's vascular surgeon dependent, I would say.
[Dr. Sabeen Dhand]
All the measurements then, are you sitting down there on your PACS and measuring the neck and this and that?
[Dr. Ben Starnes]
Yes, we do it all. I try and tell the residents, especially in a rupture setting, if it's an intrarenal rapture and the patient is a suitable candidate for EVAR, I'll tell them really the two most important measurements are the D1 and the L1, so the diameter of the aorta just below the renal arteries. Then the length from the renal arteries to the aortic bifurcation. The limbs you can sort out later, but those are the two most important measurements.
[Dr. Sabeen Dhand]
Very good point. Obviously, we've identified like you said, a pretty antiquated system of what goes on. After the repair, then who's being notified? ICU hospitalists, you or a resident are calling them?
[Dr. Ben Starnes]
We're basically walking with the patient up to the ICU and doing a handoff to the ICU team.
Listen to the Full Podcast
Stay Up To Date
Follow:
Subscribe:
Sign Up:
AI Recognition of Emergent Aortic Injuries
At Dr. Ben Starnes’s practice, AI is starting to play an important role in diagnosing aortic aneurysms. Through the Viz.ai system, which leverages machine learning algorithms, the time for aortic injury workup can be reduced, from 45 minutes to just a few minutes. This expedited workflow is a game-changer, particularly in critical aortic cases where every minute counts. Furthermore, the platform enables prompt and effective communication across a multidisciplinary team, including ER, neurology, ICU, and hospitalists, expediting life-saving treatments. The platform's accuracy is remarkable with minimal false positives and negatives reported. Finally, the easy access to rapid and high-quality imaging, even on a smart device, streamlines decision-making, making the platform a vital tool for modern healthcare systems.
[Dr. Sabeen Dhand]
Which is pretty much the first time they're hearing about the patient and among their many other patients. Let's circle back now. When we talk about AI. Like you said, AI is a term that's thrown out by a lot of people, a lot of people who aren't understanding of tech and stuff, but AI in this setting, if we're talking about Viz.ai or other platforms that are out there, is basically a software that helps you integrate the images and the teams together. Now that you're incorporating this into your practice and you've seen it at work a little bit, is AI recognizing the aortic aneurysms, all of them, or is it recognizing the ruptures? Is it identifying them?
[Dr. Ben Starnes]
Yes. The thing that I'm really excited about in using Viz.ai which we're just now starting to incorporate into our system, is that we learned a lot from the stroke care service line where it would typically take between 45 minutes and an hour for the workflow to be successfully completed. That can be reduced to just mere minutes. What I'm excited about with the treatment of our patients that harbor aortic aneurysms or aortic dissections, is that those diagnoses can be made. Then the information can be rapidly disseminated to the team to connect the patients with lifesaving treatments like an EVAR or an open repair.
[Dr. Sabeen Dhand]
Exactly. That's what we're doing at stroke. From the second the embolus is detected or the profusion abnormality is there, I already know about it. I'm able to communicate with ER, neurology, ICU, hospitalists, all at the same time. Then even other people like the stroke nursing team and everything. That's where the communication helps and identification. How accurate is the platform? Have you seen it in action as far as identifying aortic? You can talk about Viz.ai in particular or any other one about is there false positives, false negatives for aortic ruptures?
[Dr. Ben Starnes]
No, it's really accurate. The other thing that we haven't talked about yet, is the ability to just have a platform on my iPhone where I could be out to dinner with my wife and get a call. I can easily and very rapidly pull up the images and just scan through them. The first time I saw this, I was just blown away at how fast it is.
[Dr. Sabeen Dhand]
Clean, smooth.
[Dr. Ben Starnes]
Exactly.
[Dr. Sabeen Dhand]
It's better than our PACS I'll tell you that much.
[Dr. Ben Starnes]
Exactly. Sometimes I'll get in front of my computer. I'm just waiting for the images to stream across, whereas, with Viz.ai I can get on my smart device or iPhone and just fly through those images and just really make quick decisions. You know as well as I do, in these cases, minutes matter.
[Dr. Sabeen Dhand]
Again, the viewing platform. We've talked about this whole thing now about images. Having a viewing platform is so important because you're making your approach and your decision based on the images, so that needs to be there and available. It's hard when you're waiting to get sent from another hospital or things like that.
[Dr. Ben Starnes]
That's right
AI for Elective Aortic Injuries
AI can modernize aortic care by detecting untreated aneurysms across volumes of medical documentation. Sometimes, aneurysms can be “forgotten”with respect to immediate health concerns but AI could address this issue by collating lists of and flagging patients with untreated aneurysms to be evaluated and treated appropriately. Through multiple iterations, the system would need to learn to balance timeliness and usability to avoid overwhelming providers with non-urgent notifications. Thus, utilization of AI could shift the medical paradigm from a reactionary to a proactive approach, potentially providing life-saving treatments to patients who are unaware of their aneurysm or those who have been lost to follow-up.
[Dr. Sabeen Dhand]
What about non-emergent? Let's shift our focus and go to elective aortas. Would you be planning to use AI to also identify non-ruptured aorta that are five centimeters or above?
[Dr. Ben Starnes]
Absolutely. Well, any aneurysm really, even the smaller aneurysms. Here's what happens and it's really a reflection of our outdated healthcare system. A patient will go to the emergency room, let's say with right upper quadrant abdominal pain and be diagnosed with cholecystitis, whether it be by ultrasound or CT scan, but oh, by the way, on that imaging study, they detected an aneurysm that was three and a half or four centimeters. Well, the patient will get treated for their gallbladder, but this incidentally detected aneurysm will not be addressed and it'll be forgotten and the patient, you can't rely on the patient to know that what was in their imaging report.
If the word aneurysm is in any report anywhere, Viz.ai can scan millions of documents and find those words and give you a list of patients who have aneurysms that have not been treated. Those patients can be captured and brought into your system and be evaluated and then tucked under the wing, so to speak, to treat once they reach a certain size.
[Dr. Sabeen Dhand]
Absolutely. We deal with that a lot where there's things exactly how you described it, where it's an incidental finding and then just forgotten about or lost to follow up. I do think that using these platforms will really capture and help and benefit patients to be worked into a system and followed. Do you imagine, like when we have stroke cases or when you have ruptures, this system notifies you, it sends you a message or it blasts a tone on your phone. Would you want that for any aneurysm or you would want something more-- I don't know. Sometimes I would feel like that would be disruptive sometimes if it's just elective, but that may still be--
[Dr. Ben Starnes]
Absolutely. I think that's one of the topics that people get afraid of. I can't imagine having my phone go off every time the word aneurysm was reported on a CT scan within our system, because my phone would basically just be going off all day long. I think you can tone it down to the point where you get a list or maybe one of your providers, your PA or your ARNP check at the end of the week, a list of patients to then be able to go in and look at their imaging and then give the appropriate referral when needed.
[Dr. Sabeen Dhand]
Sure. One interesting, we find this conundrum in interventional radiology, sometimes in radiology, is say you have this finding and how do you then follow up? Do you wait for the primary person to call you and make the referral or do you kind of be more proactive and you call them and say, "Hey, look, this is what's here. Do you want us to take care of this?" How do you approach that?
[Dr. Ben Starnes]
Sabeen, that's the scary part to me is that, right now the way the system is without artificial intelligence, it's a clunky outdated system where we don't have any idea who's out there that has an aneurysm or dissection that is not symptomatic. We have no idea, and we rely on the primary care providers to put in the referral to us where we can actually evaluate the images. I think by having those patients be identified across a vast, incredibly enormous system or healthcare system, we can be proactive about providing life-saving treatments to patients who don't even know that they have an aneurysm or may have been lost to follow up.
Podcast Contributors
Dr. Ben Starnes
Dr. Benjamin Starnes is a vascular surgeon at Harborview Medical Center, a Professor of Vascular Surgery, the Chief of the Division of Vascular Surgery, and Vice Chair of the Department of Surgery at UW School of Medicine.
Dr. Sabeen Dhand
Dr. Sabeen Dhand is a practicing interventional radiologist with PIH Health in Los Angeles.
Cite This Podcast
BackTable, LLC (Producer). (2022, December 12). Ep. 271 – How Can AI Help with Acute Aortic Emergencies? [Audio podcast]. Retrieved from https://www.backtable.com
Disclaimer: The Materials available on BackTable.com are for informational and educational purposes only and are not a substitute for the professional judgment of a healthcare professional in diagnosing and treating patients. The opinions expressed by participants of the BackTable Podcast belong solely to the participants, and do not necessarily reflect the views of BackTable.