false
ar,zh-CN,zh-TW,en,fr,de,hi,it,ja,es,ur
Catalog
Bifurcation Cases With OCT and Coronary CT-OCTOBER ...
Computational Models in Bifurcation PCI
Computational Models in Bifurcation PCI
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
I think our next speaker also actually needs no introduction, but I'll try anyway. So Ioannis Tsitsis is a professor of cardiology at the Middle School of Medicine, University of Miami. He is the chairman of the SCI bifurcation club. He's also very active in the European bifurcation club, and he will enlighten us on how we can integrate computational models in our practice to improve outcomes with bifurcation PCI within his next talk. Thank you. Thank you. Thank you, Beamer. Thank you, everybody. Esteemed panelists and attendees online, and thank you, Dr. Rapp, for this huge initiative here. It's a wonderful series of webinars. I think I'm going to say more or less the same thing that Dr. Collette said, but using invasive interventional imaging approaches and a little bit of simulations to show how our PCI will look like if we follow the prediction. These are my disclosures, and the whole idea, agreeably, is to do our very best to improve this problem here, the adverse outcomes associated with PCIs. PCIs help a lot, but also they come with a price of adverse events, and sometimes when it comes to left main, patients with diabetes, it can go up to 30% event rate within one or two years post-PCI. The best way to improve the outcomes after interventions is to achieve better planning, as Dr. Collette said. By failing to prepare, we're preparing to fail, as said by Benjamin Franklin. What is missing, essentially, in our approach is an individualized integration of anatomy and physiology, and AI here comes to play an important role. It can leverage all this individualized anatomical and biological and physiological data in a very seamless way to help us plan in a time and cost-effective way. With AI, we can achieve automated lumen segmentation of intracoronary imaging, or any kind of imaging, in real time, at the point of care, ad hoc, in the cath lab, co-registration with angiography, because we operate on the angiograms, not on IVOS or OCD consoles or CT scanners. 3D reconstruction, our brain is advanced, and as already mentioned earlier, we are used to operate in 3D, and it's a violation of our skills and abilities, natural abilities, when we're asked to take the 3D into 2D luminograms. FFR, very important to know the significance of the lesion we deal with. PLAC, material segmentation and characterization based on imaging. And of course, simulation, to show us how the PCI will look like before it happens. Let me walk you through these concepts. We have developed this digital STEM platform, which integrates any kind of imaging, it's very versatile, any IVOS, any OCT, plus angiography, even invasive imaging with CT. It creates digital twins of the heart arteries with faithful representation, replicas of the lumen, the wall. The biology is there, is integrated by imaging, we integrate into the 3D reconstructed wall the materials of this particular individual, and then we plug everything into the computational simulations where we add patient-specific, very realistic STEM designs and balloons, and we come up with planning of procedures. We have tested the feasibility of this approach in a series of patients, and I'm going to show you today a patient-relevant example. This is a 60-year-old woman with ischemic adenopathy, history of stents in the LED in the past, and circumflex in diagonal, and severe distal left main angiographically by IVOS, 4.9 square millimeters of distal left main, shown here in the long view. And what we did in that case is we stopped there and we took the artery, this distal left main, from the body to the computer. You see here the faithful digital twin, the replica of the left main anatomy with the distal LED, with the LED stent from the past, the MLA 4.9, the distal left main, and we went one step further. We integrated realistically in 3D the material distribution of this particular bifurcation using OCT imaging integrated to the angiogram. So fibrosis here, calcium in 3D, as shown already by CT by Dr. Collet. And we went one step even further, and we translated this material distribution into stiffness, because at the end of the day, stents don't care about the material, but they care about the stiffness they are expanding against. And you see here a very comprehensive mapping of the stiffness along the 3D reconstructed by intravascular imaging diseased left main bifurcation, and you see here the correlation of each OCT frame on this stiffness mapping. You see here this particular case. Again, distal diseased left main looks stiff. Red means very stiff. On the white side means less stiff. And you see here the flow dynamics, the acceleration of flow through the diseased distal left main, LED and circumflex. And the beauty is now we allow the AI and simulations to recommend for this particular anatomy what's the best stenting strategy, how we predilate this lesion, how we stent, what stent length, size, diameter. One stent technique here, post dilatation. And this gives us an adequate result. And the precision is very high here, because you see that the machine, the AI system recommends a minimum overlap of millimeter with the previous stent, and also protrusion of the stent into the aorta by a millimeter. So very high precision coming from integration of existing imaging, which we do every day, but maybe we don't squeeze the juice of this lemon enough. And you see here the distal left main hemodynamics before and after the virtual PCI, where you see restoration of flow in the diseased distal left main to the LED, and adequate flow at the osmo, the circumflex, mitigating essentially the need for a second stent. And again, so far, we're all simulation, all digital. And now we take this recommendation into the cath lab. And what I show you today is an offline process. But we are having data, and we're working very diligently to make this real time. And I believe we'll have this very soon, to have the planning in real time. So here we execute exactly the same plan to the dot, same balloon, same inflation atmospheres, pre-dilatation, step two, crossover stenting with a 3-5 megatron from the old stent in the LED all the way to the ostrum of the left main with small protrusion into the aorta. And then finally, post dilatation with a 4-5 balloon at 18 atmospheres, as predicted, as recommended by the AI in the simulations. And you see here the baseline before stent. And we see here the all simulation AI guided plant, left main PCI, with the result, angiographically looking great. And the almost times three increase of the diseased left main from 4.9 to 16.4, you see here also on the long view. And this is also very exciting to see the precision of the AI guided PCI compared to reality. Here is the MSD across the length of the stent. Black is the actual achieved diameter of the implanted stent by Ivos. And red is what the simulation and the AI predicted. And you see here a very high agreement between the prediction, the AI-based simulation, and prediction with the actual reality, which means that in reality, we were able to faithfully execute the predicted plan. So where we head? We're heading in the years, hopefully, near future. As we discussed already, invasive or maybe non-invasive imaging looks more powerful with the advancement of technologies lately. Use AI to diagnose obstructive CAD that needs therapy. Then use computer vision to devise a patient-specific treatment plan for this particular individual. Then plug in the robot to achieve, to deliver this recommended plan using the AI-based diagnosis of CAD coming from non-invasive or invasive imaging. And this way, we can achieve consistency versus interoperator variability, which we see a lot. And hopefully, with this approach, we can have better guidance to the operators who will still be there, guards of the operation, but being more focused and more prepared, more informed for the procedures, improve cath lab efficiency, patient satisfaction, healthcare costs. And hopefully, the holy grail of what we do is improvement of clinical outcomes. And that's the planning part of computer simulation AI. Also, it can play an important role in regulatory approval of devices. We have different techniques or different devices. We can have the so-called in silico virtual clinical trials, where we use heterogeneous, very diverse patient data, not actual patients, and extract simulation outcomes and points, which are highly predictive of actual clinical outcomes. For example, stent expansion is very highly predictive of the stent stenosis. And this approach can guide us towards very focused, very informed actual trials with the real patients, real outcomes to inform academia, NIH, industry, and so on, on the devices we have to help essentially the process to advance new scientific knowledge in a very time and cost effective way. And that's something that FDA prioritizes nowadays, because this approach, the in silico clinical trials, facilitating actual clinical trials can help us with impactful scientifically knowledge in a very time and cost effective way. We learn more in faster and cheaper way. And then third application of simulations on top of planning of procedures, on top of virtual trials to accelerate the regulatory approval of devices. We also have these tools, these techniques to help us with education and training, to help our trainees learn greater, faster, better using simulations and external reality. You see here, this is our fellows before they go to the cat lab. They wear the headsets, spend like an hour in the replica, virtual replica of our cat lab. They can play around with the eye to see how in each projection the coronaries look like, no x-ray, no contrast. Seeing is believing. Same here from our advanced fellows from the outside, from the inside. They can grab the stent, like being totally immersed in a site to see how a provisional stent looks like from the inside, from the outside, how the provisional technique interacts with the site branch. For more advanced techniques here, you see this decay culotte. You see in this case, this is patient-specific. There's a realistic lumen and wall. The stent design and material is realistic. You see here the first stent, pot now, and now you see the second stent coming in and see how it interacts with the other stent. Indispensable cognitive experience that helps us understand things better and remove the old-school teaching method, which is by trial and error on our poor patients. So we essentially moved to a totally digital era in cardiovascular intervention, in general, in medicine, where AI, simulations, virtual reality will play a central role in, first of all, democratization of expertise. Imagine what we can do with CT or invasive imaging. It doesn't matter. But what matters is that in India, in Greece, in Latin America, across the globe, we can have access to the latest for the expertise versus relying on specific experts in specific geographical regions of the world. We can use all these technologies to plan our procedures ahead, and by that, we can improve our outcomes. We can use these technologies to develop faster and more efficiently our devices, virtual clinical trials for accelerated pathway through FDA, and of course, education and training to help our trainees learn greater, faster, and better. The future is exciting and it's happening. We had the honor to put together a compilation of many papers, review and original papers, in this J-Sky special issue that just came out a few days ago. I had the honor to co-edit this with Dr. Edelman from MIT. Take a look. I think it's worth it to read these papers, these articles. A big thank you to the members, my group at the Center for Digital Cardiovascular Innovation at the University of Miami, engineers, MDs who work hard in these concepts. Of course, a big thank you to the funding sources from the NIH, from industry and philanthropy, and a big thank you to all of you for your attention. Thank you. Thank you very much. That was an amazing presentation. If I could just ask one question. I think you actually have the potential to lay another layer on top of what Carlos Collette just presented in his previous presentation because not only maybe will we be able to predict what is the FFR after the procedure, but having these newer computational fluid dynamic type of analyses could also help to perhaps guide us which type of bifurcation technique would be optimal to get optimal flow characteristics after a procedure. Obviously, doing this offline would be easy. I know that computational fluid dynamics takes a lot of computational power as well. You need a machine to run for a couple of hours to get these models right. But if you're able to do this perhaps with some advanced AI technology to get it integrated into the CAD lab, that would be an amazing step to maybe even help you while you're doing these procedures and maybe switch tech if you think that a different approach may have some benefits. Could you maybe comment on the developments that are ongoing in us using these techniques real time in the CAD lab maybe in the near future? Without question, Bimmer, interventional cardiologists would feel very unhappy if you had to take a patient off the table, make a planning, and bring the patient back. This would be totally unattractive. Count my words, what we do at this point and where we're heading is real time anatomy, physiology, biology, and planning of procedures at the point of care ad hoc in the CAD lab once we hit the IVUS pullback, the OCD pullback, within a matter of one or two minutes to have the whole operation planned again in real time. That sure sounds exciting. So John, you're a budding imager, inter-corner imager, so what are your thoughts? I mean, I think that's fantastic. I think one of the things that I've loved about OCT is the prescriptive nature of MLD MAX and walking in with a plan. So anything, and to your point, the efficiency of it, right? You're not guessing, you're not trying different things, you're imaging and working through an algorithm. So I think if the AI has the ability to give you that in one or two minutes, it's a game changer. It's really impressive work and a really nice presentation. Thank you. So for all of you, this is available on Sky TV to look at. One thing, so Ioannis, for your case that you showed, why didn't you do a follow-up imaging with OCT rather than IVUS? We started with an OCT-guided three-dimensional acquisition. So why did you not end with the OCT follow-up in that case that you showed? It's a matter of preference. We tried, we started with OCT. We ended up with IVUS. I don't think that HDIVUS is also very high resolution. We could do OCT. It's, we just elected to use. You could do the pre-planning with HDIVUS as well. Is that correct? Of course. You can do pre-IVUS, post-IVUS. You can do pre-OCT, post-OCT. You can combine them. You can do CT. You can do CT. Again, we have plenty of time with CT to do the 3D, essentially what Carlos showed. And then on top of this, do the simulation. Show how this provisional standing in this trifurcation would look like before you actually do it. And this will give you important information to get a sense, a cognitive understanding of how the stand would interact with the side branches, how they would look like. Again, we talk about patient-specific models which integrate anatomy, physiology, and biology. So there's a question for Carlos. Carlos, what about aortic osteolesion in your algorithm, or a 010 lesion in your algorithm? So aortic lesions, for me, is one of these lesions subset that you don't want to do anymore without CTs once you get used to that. And the reason is you can really understand what is the precise angle to really see the osteoporosis perpendicularly. There's a paper published this month in Circulation Intervention showing basically CT guidance for osteoporosis PCI with the end point is Ivo's geographical niche, angio versus CT. And you see that when you use CT, you land where you wanted to land, basically, because you were guided which projection was the best one to basically show you what you wanted to see. So I'll tell you a personal comment. When I go to the lab today, and it's an osteo lesion, and I fully agree with you, Janice, what you said, that you don't want to take the patient off the table and then bring it back. But there are a couple of scenarios where you don't have a CT and you don't know exactly where you are. Having this information from CT really makes the procedure safer for the patient. And one of these substances is osteo disease. So I think osteo disease, bifurcation lesions, CTO, are sweet spots for CT planning. Well, we're coming to the end. But you know, we have to get a respect imaging altogether. And I think that is a big call in the US. We still have problems with insurance, as I said. But I think this is the way to go. But even photon CT comes along. And I think there's a lot for the younger generation intervention card. Hopefully, you can integrate this all into your practice. How do you get away from the radiology turf, Carlos? Radiology is trying to be involved in this. And we are interventionists trying to take this away from them. How did you manage that? I mean, this is a big issue for us. Yeah, that's a good question. So I think with our radiologists, we work together. So basically, we look at the tough cases. We do that once or twice a week. We meet and we see the cases. What we have seen is that the collaboration has really increased. And we try to teach them PCI for them to report what we wanted to learn. But at some point, they said, guys, we will never learn what is relevant for PCI. So it's better that you get involved. And then we'll basically share the program. So that's the way that things evolve from our side. Thank you very much, Bimmer. Mahesh, any final words? I also think with the CT, the way in Europe, we have these hard teams where we work very closely with our radiologists as well. When we do our SAGI team, the SAVR team, the radiologist is also in the room. Well, I know in the US, there's a lot of planning being done by the cardiologists themselves or by the companies, for example. So really, the way to be able to get CT to the next level is to collaborate with your radiologist on the next level. And I think in Europe, we've been doing that, particularly in the SAVR space for a long time. And the coronary space, now that it's evolving and now that the software is there and the AI integration is there, we're seeing great benefits from that. Because it's just that much easier to extend it into the coronary field that the collaboration that was already there. Marsha? Thanks. I think two incredible presentations showing what a lot of tools we have out there to improve outcomes. And I suppose the only caveat, if at all, would be integrating this into training. Because already, things like intravascular ultrasound, most are under-trained. It's not incorporated in many programs of mandatory fellowship training across the globe. So integrating all of this into training would probably be the next challenge. But nevertheless, it's two incredible talks. Thank you. Thank you very much. Time has come for us to stop. Thank you, Laura, for being involved today. And we hope to see you all soon. Thank you very much. Excellent presentation. Thank you.
Video Summary
The presentation by Professor Ioannis Tsitsis, a cardiology expert, highlighted the integration of computational models and AI in enhancing percutaneous coronary interventions (PCI) for bifurcation lesions. Tsitsis emphasized the potential of AI in optimizing procedure planning by providing real-time anatomical and physiological insights, reducing adverse outcomes such as high event rates post-PCI, especially in complex cases like left main lesions and patients with diabetes. The digital platform they developed can create digital twins of heart arteries, facilitating procedure simulation and precise stent placements before actual surgical interventions. This approach can democratize expertise, improve clinical outcomes, enhance cath lab efficiency, and significantly benefit patient education and device regulatory processes through virtual trials. Tsitsis also addressed the necessity for interdisciplinary collaboration, particularly with radiologists, to leverage advancements like CT imaging to optimize intervention strategies. This innovative integration aims for consistency, reduced variability, and fostering a highly informed procedural approach.
Asset Subtitle
Yiannis S. Chatzizisis, MD, PhD, FSCAI
Keywords
AI in cardiology
percutaneous coronary interventions
digital twins
bifurcation lesions
interdisciplinary collaboration
procedure simulation
×