Selected Podcast

How AI Is Revolutionizing Spine Surgery

Northwestern Medicine’s Srikanth N. Divi, MD, an orthopedic surgeon, and Najib El Tecle, MD, a neurosurgeon, discuss the transformative impact of AI on spine surgery. They explore how AI enhances diagnostic accuracy, reduces complications and improves surgical outcomes. Additionally, Dr. Divi and Dr. El Tecle delve into the critical aspects of AI in medical applications, including machine learning and imaging analysis.

Learn more or refer a patient to the Northwestern Medicine Center for Spine Health 


How AI Is Revolutionizing Spine Surgery
Featured Speakers:
Srikanth Divi, MD | Najib El Tecle, MD

Srikanth Divi, MD is an assistant professor of Orthopaedic Surgery and Neurological Surgery at Northwestern Medicine. 


Learn more about Srikanth Divi, MD 


Najib El Tecle, MD is an Assistant Professor of Neurological Surgery at Northwestern Medicine. 


Learn more about Najib El Tecle, MD 

Transcription:
How AI Is Revolutionizing Spine Surgery

Dr. Will Meador (Host): This is Better Edge, a Northwestern podcast for physicians. And I'm your host, Dr. Will Meador, Associate Professor of Neurology at the University of Alabama at Birmingham. Today's topic is Artificial Intelligence in Spine Surgery. And today, we have joining us Dr. Srikanth Divi, an Assistant Professor of Orthopedic Surgery and Neurological Surgery at Northwestern Medicine; and Dr. Najib El Tecle, an Assistant Professor of Neurological Surgery at Northwestern Medicine.


Dr. Divi, AI use is exploding everywhere and it's a hot topic. How is AI use relevant in healthcare operations and can you highlight the relevant AI terms as well as the impact of public access to AI?


Dr. Srikanth Divi: Sure. You're right. AI has exploded, especially in the last several years. It's now pretty mainstream, and I think a lot of people are familiar with the term. Just to kind of back up and define what AI actually means, you know, the technical definition is something like a machine that can replicate human-like intelligence. So, it has a pretty broad definition. But when we look at it, you know, within the realm of whether it's medicine or something technological, we're looking at specifically, one, a few certain subsets.


One of that is machine learning. You might have heard of that before. Machine learning deals with complex mathematical and statistical models to create outputs. And then, deep learning is maybe what might be more familiar to the layperson. That's where these models have the opportunity to improve upon themselves and create what's kind of like a neural network to more accurately predict outcomes. And then finally, um, applying all of these towards text or images is probably how we think about it in the current day, where we have what we call generative AI. And this is where, you know, we can type in a question to ask a chatbot. We can ask it to create an image. And that's where technology is moving towards nowadays.


Host: And Dr. El Tecle, what are aspects or features of AI that are crucial to medical applications?


Dr. Najib El Tecle: So, given the definitions we just talked about, there's the question of how can this new computer capability help us in the setting of healthcare. And in healthcare, as you know, we're trying to optimize patient's experience, patient outcomes. So AI, hopefully, was all that data that we're building into it is going to help us on that journey. We can use the data sets we're collecting to try to decide what surgery patient is going to get. It can help us predict how a patient is going to do after surgery, but also more so it can help us optimize how we do spine surgery. For example, for the specifics of spine surgery in a cost-efficient way, how do we help distribute the limited resources we have access to in a more equitable fashion, how we can achieve or be better stewards of the resources that we have.


Host: And Dr. Divi, can you speak to how machine learning specifically can help spine surgeons, highlighting that importance in medicine?


Dr. Srikanth Divi: Yeah, I think in the context of like clinical care, machine learning can be pretty helpful. We have, especially in Surgery, a limited cohort of patients to study from. You know, whereas in non-surgical specialties, there are much, much bigger pools that you can extrapolate, outcomes.


So, with machine learning, if we're able to combine resources with bigger data sets, we may be able to more accurately predict who's at risk for a complication. What is the best, you know, method for doing surgery? This can, in the long run, improve patient outcomes, improve their safety, decrease costs by, you know, decreasing the number of complications and improving patient time in the hospital. So, there's lots of factors there. where machine learning can directly impact spine care.


Host: And Dr. El Tecle, thinking specifically about the application in Surgery, how can AI enhance a surgeon's ability to diagnose or plan a treatment or even perform a surgery?


Dr. Najib El Tecle: That's a very relevant question in the setting of spine surgery. You may have heard the saying that no one gets one spine surgery, right? And that in a way highlights like the stuff we don't know about spine surgery patients, right? The hope is that AI is going to help us optimize whatever we do for the patients so their risk of needing more surgery down the road is diminished or minimized as much as possible. Now, spine problems are part of aging, and there's no way we're stopping aging. But if we do the right surgery in the right setting, maybe we can slow down that process.


Host: And Dr. Divi, getting back to the comment about improving patient outcomes, can you explain more specifically how AI can enhance or improve patient outcomes in spine surgery?


Dr. Srikanth Divi: Yeah, so, There's several different ways we can address that. I think one of the first ways is patient selection. So, a lot of times we do this work inherently. When we see a patient in clinic, we look at their, you know, comorbidities. We look at the actual spine pathology. We weigh the risks of what kind of surgery we suggest, you know, are they able to handle surgery? And we try to, you know, infer what the outcomes may be. What machine learning can do is use, you know, a vast data subset and more accurately predict the patient outcome. As we know, humans were pretty susceptible to recall bias. We're pretty susceptible to lots of other things that may directly, you know, tamper with what we think may be the best outcome.


And then, finally, you know, each surgeon is going to have a different way of looking at things. One way machine learning can help is also kind of standardize a sort of treatment. Now, you don't want to standardize too much because every patient's different, but that's where machine learning, which incorporates so many variables, you know, it can introduce more granularity than was possible before.


Host: Dr. El Tecle, thinking about predictive models or imaging analysis, how can AI help to reduce complications in that manner?


Dr. Najib El Tecle: In the setting of imaging analysis or predictive models, what we are trying to do is input into an AI model what the current patient condition is and try to predict what's going to happen. And yeah, we probably have been talking about this a couple times that there's a recurrent theme here. We're using a model. It's helping us tell what's going to happen down the road in a more precise fashion.


When we try to analyze the images of a patient before surgery, for example, a common topic in spine surgery is how is the patient standing up? Do they have scoliosis? How are they leaning? Are they stressing their spines in unusual ways? Now, newer models that here we're doing with industry partners are helping us predict how is that patient going to stand up after we do the surgery. Things like that where in the past you had to rely on your imagination. And your imagination, you may have gotten it wrong with your imagination.


Now, we have more ways of saying, "Look, in a way If that model that has seen more like 100 patients with a very similar condition and that was the outcome, probably what you're doing here, you're going to get a similar outcome to those. And you can readjust your surgical plan even before you enter the operating room." And this is where planning properly with the help of AI hopefully augments surgical outcomes and allow these patients to have a better postoperative course, because we planned what things are going to look like for them before we started the surgery.


Host: That's great. And Dr. Divi, thinking about that comment of predicting how people might stand after a surgery, how can AI be utilized to optimize alignment in spine surgery to make it successful, thinking about decompression goals or nerve pain and how that relates to the surgical outcome?


Dr. Srikanth Divi: Yeah, that's a great question and you know, to kind of break it down for listeners who are not spine surgeons, there are a few factors we think about when we do spine surgery. Number one is neurologic decompression. Get rid of where the problem is. Number two, think about the alignment. And if we're doing a fusion surgery, that becomes pretty paramount because then you permanently fix the patient like that, and long-term, if that's not optimal, then they can have, you know, the revision spine surgeries down the road that we're trying to avoid.


So, what AI can do is, by analyzing how patients have done in the past, suggest a surgical approach. For example, you know, we know that coming in anteriorly or laterally can get better alignment goals. A lot of times, we like to do surgery posteriorly and that may not be enough. And so, this is where, you know, the model can suggest one route or the other. And then, in the context of the patient-specific comorbidities, for example, not every patient is a candidate for an anterior or lateral approach. So, that's where the model can help guide us. Obviously, this is something we do on a daily basis, but this is where the model can be sort of a co-pilot and help us guide us along the right path.


Ultimately, we have to be the kind of fact checkers and just make sure that we're doing the right thing. But if we want to achieve the optimal alignment, then, you know, we could start with what, let's say, this copilot says, and then go from there.


Host: And Dr. El Tecle, earlier, we discussed that there is the possibility for AI to predict kind of outcomes. As far as revision rates and complications, do you see that there's going to be an effort there to utilize AI to predict who's going to have complications from these surgeries, and maybe predict who's going to be most successful?


Dr. Najib El Tecle: Absolutely. So, we currently have statistical models that will tell us, "Look, this is a patient at a higher risk of complication." We even here at Northwestern set up a system so that these patients who are at higher risk are either avoiding surgery or getting optimized to get certain variables under control. And then, they get their spine surgeries. Now, the hope with AI is we, one, we get more precise. Two, we're able to use that and leverage it to allow these patients to do better.


Sometimes, for example, certain variables may not be as important as others, and we may be holding surgery of these patients because they don't meet all the criteria. Well, maybe they don't have a higher risk profile, or maybe we can optimize a few things to make things better for them, and they become surgical candidates. I think what AI is going to do is it's going to take those basic statistics that we know, sure, if you smoke, your chance of having a wound complication is high, and it's going to become more of a comprehensive picture of the patient. Like we can look at the patient chart, maybe even AI can look at the patient chart, look at their images and be like, "Oh, look, be careful. This patient's chance of complication is higher." Now, how do we act on that? Do we take additional measures to prevent it? I feel that's a question that I'm not sure what the answer to it. I don't know. What do you think?


Dr. Srikanth Divi: Yeah. I think if we're using AI as a way to optimize patient selection, that's, you know, what it's intended for, currently. I think, as we gain more patient data with larger data subsets, that can become more accurate. I feel like we can split this, you know, with application of AI into like several different ways. One is patient selection. And one is like the imaging analysis kind of alignment goals that we talked about, which then ties into the outcomes. And then, there's also like intraoperative technology, which also ties into patient safety and outcomes. It's really hard to, you know, separate out all of these variables. But in tandem, as we progress, I think we'll overall go to a swing in the right direction and improve patient safety and outcomes.


Dr. Najib El Tecle: And I feel with these bigger data sets, AI is data hungry. The more we collect data, the more accurate it is, because if you take Dr. Divi's patients, my patients, still there's one confounding factor that is I'm the surgeon, he's the surgeon. Once these data sets become way larger and they include patients from throughout the country, throughout the world, then I believe these models even become more accurate, as you're saying.


Dr. Srikanth Divi: Yeah, I definitely agree with that.


Host: And obviously, Spine Surgery and Radiology go hand in hand. It's so crucial to have good information from radiologic interpretation. So, how do you see AI interfacing with radiology reporting to assist spine surgeons, Dr. Divi?


Dr. Srikanth Divi: Yeah. Well, one thing that we're actually working on right now is a research project just to see is how can a large language model translate a radiology report. I don't know about you, Najib, but I get patients on a daily basis in clinic ask me what this word means on a radiology report. And oftentimes, it has nothing to do with what they're coming in for. And so, you have to spend time explaining it to them, you know, this is what this term means and it's just a difference in language basically.


But when we're doing this project, we're running the radiology report through ChatGPT and it's nice in the sense that it simplifies the problem into lay terminology. It says, "Look, your problem is at L4-5. You have instability, and it's going to be pinching the nerves there. This is the problem area. Don't worry about all these other things that are on the report." And that I think can give a lot of comfort to a patient, right? They're not worrying about this, what's this word up at L1-2? And I think it can provide a lot of reassurance. So, you know, it's a radiologist's job to look and report everything so that that's what they do.


In this case, that's one instance of how AI and radiology go hand in hand, and that can help. There's other things that other departments here at Northwestern have done too, looking at, you know, for example, incidental findings on chest CT and, you know, whether those require workup or not. So when looking at actual imaging, AI can analyze imaging and determine whether something needs to be worked up further or is more benign. And then, with reports themselves, you know, maybe certain terminology can be flagged so that, for example, maybe not necessarily in the spine surgeon's clinic. But if it's a person coming to a primary care doctor or in the emergency room, if physicians that don't, routinely interface with spine patients, they may know who to refer to more urgently.


Host: Yeah. And I think that lay language interpretation would be key for myChart messages and portal messages to make it clear to patients what they're reading, that would be tremendous as well. That's absolutely right. We've heard a lot about the benefits to the surgeon from AI and that technology. Dr. El Tecle, what do you think about the patient experience or patient education applications of AI in spine surgery?


Dr. Najib El Tecle: Yeah, I want to elaborate a little bit more on to what we were just saying. That is like, it's fantastic what's happening here, right? This is democratizing healthcare. This is bridging the healthcare literacy gap where patients who may have no idea what all these words in the report are saying are now getting that report or that sentence that summarizes what's happening. And now, it's easy to take that sentence and ask more questions and try to understand really what's going on with their health. I think that's really fantastic Yes, AI can help with one physician burnout too because now you're not getting all these myChart messages, and there are currently applications where AI is templating your response to a myChart message. You just have to double check it and be like, sure, that makes sense. And that's the answer I want to send to that patient.


I think all this happening makes me really excited about the future. In a way, I just started two weeks ago, using this AI scribe that we're traveling here at Northwestern, and it's fantastic. Like the way it affected my workflow, I really think there's like a before and after. So, I'm really excited for the future of AI in medicine in general, but also for us as spine surgeons.


Host: And Dr. Divi, what should physicians be most excited about the potential of AI? Obviously, you know, reducing messaging or pre-templating notes is fantastic, but what are other benefits of AI?


Dr. Srikanth Divi: Yeah. So for physicians, let's say example, let's take a spine surgeon, so directly for us, how this can help us is what we're talking about as a decision support tool, which patients are surgical patients, which patients are not. And in our clinic, obviously, we want to see surgical patients that we can help. if we can't help them with surgery, then we want to know who to refer to.


Oftentimes, we see a lot of patients where there's a mismatch in their symptomatology versus their imaging. And then, it's our job to figure out, "Okay. Should they, for example, you, a neurologist? Should they see a rheumatologist or someone else to get further workup because there may be some other systemic cause that's going on?" So, that may be where like a decision support tool can analyze their prior records, their prior tests, and suggest what the next step is. Maybe there's a, you know, a risk for MS or some other condition. So, that's from a spine surgeon's perspective.


From an overall broader healthcare perspective, I think we are, as a country, asking more of less providers. And, you know, with an aging population, we know that this has been cited in kind of every aspect of healthcare. There's going to be more demands on the system. So if we have an accurate algorithm or application that can, for lack of a better words, triage patients or just know who needs more urgent care or maybe needs direction to the right care, that's where I think AI can help. So, it can kind of fill the gap, so to speak, in terms of the demands for the healthcare provider.


One thing we're working on here at Northwestern that recently got some grant funding for is to create a sort of a digital consultant. So, for example, someone comes in with a cervical spine problem into our ED. You know, we want to be able to analyze all the text reports of the ED physicians, the physical exam. And then also, any prior internal medicine or other doctor's notes, and then analyze the text reports from their imaging, and then sort of have the application tell us, "Okay, here's like the likely differential diagnoses. Here's like a rank order list of what the next steps should be. Does it need more imaging, or is it benign and just needs to follow up in clinic? Or should we, you know, admit, and do surgery?" So, I think things like that can really help. We're privileged here to have an excellent set of physicians in every subspecialty, but, you know, what happens if you're practicing in the middle of, you know, rural setting and you don't have access to that? So, this can really help those kinds of physicians.


Host: Absolutely. And Dr. El Tecle do, you want to provide us with some key takeaways for the providers listening today on the discussion of AI and the potential and future directions.


Dr. Najib El Tecle: I think the future is really exciting. I really like sometimes try to imagine what will medicine look like in 2030 and what will spine surgery look like in 2030. And I think that change is happening as we speak. I cannot overstate the impact AI is having, and is going to have in the future. It's really great to be practicing spine surgery in 2024.


Dr. Srikanth Divi: Yeah, I'll piggyback on that and say I completely agree. It's improving, or has the potential to improve every avenue, whether it's in the OR, outside of the OR. So, I think we have a lot to look forward to.


Host: So, this is all very exciting information. And when we want to get patients to you for evaluations at the Northwestern Medicine Center for Spine Health, how should we do that to connect patients to your clinic?


Dr. Srikanth Divi: To connect patients to our clinic, we have a call center where they can call in, and they can be routed to the specific provider based on their symptomatology. In our spine center, there's nine surgeons currently, six neurosurgeons and three orthopedic surgeons. So, we all act as a collegial group and we all see patients together. So whether it's through a direct referral with, you know, a provider or just coming in through the call center, there's information available online how they can contact us.


Dr. Najib El Tecle: And if I may add that Spine Center, we also offer physiatry, physical therapy, and pain management as are some of the options that are available within the Spine Center. That's right.


Host: Well, thank you both for your expertise, and this is really exciting stuff. Looking forward to the future of spine surgery here. And remember to refer your patient or for more information, you can head over to our website at breakthroughsforphysicians@nm.org to get connected with one of the Northwestern Medicine physicians. That wraps up this episode of Better Edge, a Northwestern Medicine podcast for physicians. On behalf of our team, I'm Will Meador, your host, and thank you for joining us today.


Dr. Srikanth Divi: Great. Thank you.


Dr. Najib El Tecle: Thank you.