false
Catalog
Efficient and Focused Cardiac MRI (2023)
S4-CCA07-2023
S4-CCA07-2023
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
So, I'll kick it off here, and the title I was given is Fast and Focused Cardiac MRI in 20 Minutes. So, if you think about cardiac MRI, how it is done today and maybe at your clinical practice, it's often perceived as very complicated and long exams with a large number of images and really complex data analysis workflow. You can see an example if you kind of, you know, pull a study from your PAC system, it looks like that. And it actually does look a little bit overwhelming and complicated, but I kind of want to kind of debunk this a little bit and actually talk a little bit about that it's actually not that complicated, and if you kind of focus on how to construct efficient protocols, you can actually do CMR in 20 to 30 minutes relatively easily. 20 minutes is a bit harder, but 30 minutes is certainly doable. So, let's just take a step back and look at the core cardiac MRI exam. And if you compare that to the slide before, it actually looks much simpler, so it's actually not that complicated. It really consists of scene imaging to look at global cardiac function, wall motion, first pass perfusion, stress perfusion, typically rest and stress, then there's the late gallium enhancement, and you may even add on some flow imaging. So, it's not that complicated. Nonetheless, if you kind of run a protocol like that, you still have to kind of deal with a lot of kind of interactions with the patients. For example, scene imaging is about 12 to 14 slices per series, so this will be about 12 to 14 breath holes traditionally. Scene perfusion, there's stress and rest, so this is two breath holes. LGE is 12 to 14 slices covering the entire heart in short and long axis orientations, so 12 to 14 breath holes, and then flow imaging as well. So, this seems kind of a lot to go through, and this kind of also explains why these protocols are complex, cumbersome, require a lot of interaction by the technologists, breath hole commands, and things. But there's actually a lot of things you can do today on every MRI system that's out there right now to really make this more efficient and achievable in shorter scan time. So, I want to kind of go through these individual techniques individually and talk about what's available today and what's on the horizon and how this could be implemented in fast and efficient protocols. Let's start with scene imaging. This is a typical scene imaging where you cover the heart in short axis orientations, and standard scene imaging has traditionally been done with eight slices, eight breath holes. So, this is breath hole instructions, you know, everything, so it takes quite a while to get through all of these slices. But you know, all of the MRI systems have parallel imaging now available, accelerated CMR is available, so you can really acquire two to four slices per breath hole, so this really would break this down in two to four breath holes. On top of that, there is really good standard real-time techniques available now where you can, in patients with atrial fibrillation or have a difficult holding of breath, actually can do these scans, scan actually with shorter exam times, even during free breathing. And there's many other techniques on the horizon out there, as shown here, that have now kind of a very intelligent approach to scene imaging, where they kind of acquire data over a longer period of time during free breathing, requires no breath holes, kind of acknowledge arrhythmias while they are acquired, and then reconstruct these data sets. An example is shown here, so standard breath holes, 10 slices, if you just record the time it takes, it takes about 200, 300 seconds, so about five minutes. But if you look at these free breathing techniques with motion correction that require much less user interaction, because they're free breathing and just free running, this can be done in about a third of the time. So these are not yet commercially available, but really on the horizon. But even with the existing real-time techniques and the parallel imaging, you can cut down your time sufficiently. Stress and rest perfusion, same thing. There's many good techniques out there that can do this in two minutes for stress and two minutes for rest. And there's even on the horizon many techniques that would do the quantification and do quantitative microperfusion on the scanner and make the data available. And I will not spend much time on that, because the next speaker will actually have an entire talk about rest and stress perfusion and how to do it as efficiently. So I'll defer to the next presentation for more details on that. And then if you take the third component, conventional late gadolinium enhancement, again, this has traditionally been a segmented acquisition. That means you have to do a breath hole over eight to 12 heartbeats. And again, if you want to cover the entire heart, many different orientation slides would be eight to 12 breath holes. But really, there have been many developments that are available routinely, clinically now that do one slice per heartbeat or are free breathing. So you can really do eight to 10 slices at a breath hole, or maybe have only two to three breath holes. So we acquire that data in a much shorter time. And on the horizon, I'll talk a lot more about this later, there's accelerated 3D techniques that are free breathing and whole heart that will make this even easier. So you just acquire data for a little bit prolonged period of time, but then you really don't have to deal with any patient interactions. And then other techniques on the horizon in LGE are really, again, these free breathing and motion correction techniques. And this is a really interesting approach where you have a highly accelerated single shot acquisition, but you repeat that acquisition multiple times during free breathing. But then, of course, the heart would move between the different heartbeats. And if you apply motion correction, you can actually take these different kind of heartbeats that kind of map out the LGE, motion correct them, and then average them, and really get a high resolution, high contrast to noise image. And this can then even be applied in a patient with atrial fibrillation, as shown here, where you can see this kind of erratic motion due to AFib, motion corrected, and then a motion corrected average image that really nicely delineates the delayed enhancement here in this particular patient. So if we go back to the standard core exam that I showed earlier, and if you kind of look again at the different components of this, the CIN imaging you can really do in much shorter time, maybe four to six breath holds, even real time or with motion correction. Same is true for the LGE. It's only two to three breath holds. There's real time flow imaging. I haven't talked about that. And perfusion is about two breath holds, and you can do this in a couple of minutes. And again, you'll hear more about this in the next presentation. So this really brings it down to a 30-minute exam that you can now actually do easily in probably 80% or 85% or 90% of the common indications for cardiac MRI. And I'll point you to this recent white paper that was published by SCMR that details how to do a 30-minute CMR exam for common clinical indications. And it really kind of takes advantage of what I've talked about there, organizes the protocols efficiently so that within 30 minutes, you can really do your localizers, your CINE exam, perfusion, 3D LGE in a coherent way for really all patients. And this is here as it's right here now. So this is not like future directions or any techniques that are not available to you. You can do this now. If you kind of structure your protocols accordingly. But of course, the talk is CMR in 20 minutes, not in 30 minutes. So we still have to shave off 10 minutes from here. And really, the fact that we're not at 20 minutes or 15 to 20 minutes routinely is really, in my mind, to two main challenges. And one is workflow. You still have to have all these interactions. And then the other one is really the need for breath holds. And those are obviously interrelated. And the way I look at it, really two different approaches or two different solutions to this. One is really not so much changing the imaging techniques, but attacking the workflow issue. And the other one is really changing the paradigm how we do imaging. And I'll kind of talk about both of these techniques in the next half of the presentation. The first one is really what's referred to as the one-click CMR protocol. And this is an interesting concept I find. So for example, you use, again, your standard type of protocol with localizers, T1 mapping, stress perfusion, scene imaging, and LGE. And this would take you about 29 minutes. But keep in mind, a lot of that time is spent on positioning of slices, breath hold instructions, and things like that. And there's many institutions and commercial vendors that are working on these automated and AI-supported workflows. In other words, you don't really change the imaging techniques, or you don't use the most kind of fancy, highly-accelerated techniques. But you kind of use AI and software solution to automate the planning and the execution of the cardiogram. The ultimate goal is really that one-click exam. So the idea is you put the patient on a scanner, click once, and then the AI takes care of scan plane positioning, executing breath hold commands, and so on. And this is actually out there. There's prototypes out there that have been used, and I'm just showing one example from the group in Libic, where they've used this data for now a small cohort of 44 patients. This was presented earlier this year at the ECR in Europe in 2023. And they looked at 44 patients with cardiomyopathy. And they used this AI automation for isocentering, positioning of imaging volumes, saturation bands, navigators, and even an auto-TI for late gallium enhancement. And what's really interesting, like in 91% of cases, so in the vast majority of cases, there was no need for user interactions. The protocol just ran through in under 30 minutes, fully automated. Three cases, they needed adaptation, and those, it's kind of obvious why, there were congenital heart disease cases. Obviously, the AI wasn't trained on those cases, and they're so complex. There's user interaction. But you can imagine, even with these user interactions, you probably save a lot of time and reduce the complexity of the exam. And in one case, the auto-TI was incorrect. So this is very promising. This is under development. But I think you still see more of that, and maybe even some hybrid versions of accelerated techniques in some AI. So we'll kind of have to pan out what the ideal solutions are there. And you could take it even further. This is an example here from Belgium, where they kind of used this AI auto-align, auto-positioning, auto-execution techniques, and combined it with highly accelerated compressed sensing imaging, and, you know, these kind of accelerated motion-corrected heart-freeze techniques. And they could do really highly accelerated scan that gave you function and LGE in under 12 minutes, or in about 12 minutes. So you could imagine, even if you add in T1 mapping or stress perfusion, you could now really talk about 20-minute CMR scans. So this is really exciting and promising, I think. And then the second approach to this is very different, and that is really a change in paradigm how you acquire the data. So if you think about what we're doing right now, we do everything prospectively. We define a slice, we do EKG gating, it's a 20-second breath hold, we define the number of time frames per cardiac cycle. This approach is completely free-running in 3D, so it acquires data all the time for 10 minutes, and it interleaves the acquisition acquiring data that you can use for motion correction, it's highly accelerated, it's multi-contrast, there's some kind of physics behind it, and then it's 3D. So the idea is really to run the data, not prospectively, just run, acquire data for 10 to 15 minutes in a smart way, and then after the scan is finished, kind of tease out the different components that you're interested in and display those. And I'll show a couple of examples. This is from Matthias Stuber's group at University of Lausanne in Switzerland. This is a so-called fully free-running technique. It's fully respiratory and cardiac gated, and it has a highly accelerated reconstruction. What's important here, this is really about ease of use. You don't need any EKG leads, no respiratory navigator, you just place a 3D volume of the heart, press start, and then collect data. And after you've done this for about 10 minutes, then you re-bin or re-sort the data along the cardiac and the respiratory dimension. And then you can flexibly tell your data set, well, I want to look at Cine over time, so you can reconstruct the data set for one respiratory phase over time. Or you can say, well, I'm only interested in end-systole, but I want to see how the heart moves its respiration. You could do that as well, or you could just freeze everything and look at the 3D image of the heart. So the idea here is, again, you don't plan prospectively what you're imaging, you acquire data for a long time, and then after the fact, you decide on kind of what you want to look at. And you can kind of enhance that with adding in additional information. This is data from the group in King's College from London, where they have a similar approach, but then they now added T1 and T2 mapping and kind of the imaging of FAT as well in eight minutes. So this is a 3D volume, not time resolved, but you can see it's, again, this acquisition. And after the fact, you reconstruct data on T1, T2, and FAT, and this is a comparison here between the kind of standard breath hold and the kind of 3D reconstructed data. Here's another example here from UCLA, from Anthony Cristodolo, who kind of had this approach where he now takes a six-dimensional approach. And that just means there's two spatial dimensions, there's cardiac motion, there's respiration as before, there's T1 and recovery, and T2 prep. So data as a part of the smart way that all these contrasts and include your data, and then after the fact, again, you kind of bin the data in different states and reconstruct it. So an interesting theory, if you do something like this, is this is now a 2D approach, not 3D, but you now get things like T1 mapping, T2, and T2 star as a function of the cardiac cycle. So you get dynamic T1 and T2 maps, and maybe you could even from those maps or data collect your CINA data. And this really only takes two and a half minutes per slice. So you can see, if you think about these techniques that I just talked about, it's really a new paradigm in terms of imaging. It's completely free-running, 3D. It's kind of early. I don't think there's anything available out there that you can use as a commercial product, but I think they're very promising. Obviously, there's a lot of post-processing that's needed, because data is 3D. You kind of need to re-slice it. But again, AI is very versatile in that area and could help a lot there. So in summary, I hope I was able to show you that efficient CMR is not something of the future. It's here today. State of the art. There are many techniques for efficiency improvements. 30-minute exam for common clinical indications is here. You can do it today. And there's really, on the horizon, highly accelerated and AI-supported CMR with existing techniques that hopefully will, particularly from a workflow perspective, make it even more efficient. And then what's really out there is the one-click protocols and these free-running 3D techniques that I just talked about in the end. And in the end, it might be a mix of both. And we'll see how the future develops there. But very promising, a lot of new approaches out there that are very exciting and to look out for. Thank you very much for your interest. So my task today is to talk to you about quick ischemia imaging with cardiac MRI. So there are some subtle and not-so-subtle differences in the types of stress cardiac MRI. So I'll begin by discussing about choosing correct protocols. This way, we can ensure we are answering the clinical questions whilst trying to rapidly scan patients. I'll then break down the stress cardiac MRI procedures so we can find time savings in these various aspects. So I'll begin by covering preparation and drug choice. I'll next discuss the scanning and protocol, followed by interpretation and post-processing, since reporting images and analyzing them is also an area which we can use efficiently. So choosing the correct protocol. So as mentioned, we want to save time, but we need to answer the clinical question. So in terms of answering the clinical question, obstructive coronary artery disease, or CAD, is the most common reason for stress CMR. Coronary microvascular disease is an increasingly common request, also with a lot of interest focused on ischemia with non-obstructive coronary arteries, also known as Inoco. Anomalous coronary arteries is another request in order to identify ischemia as a result of usually the malignant course of the anomalous coronary artery. Lastly, under specific circumstances, cardiologists sometimes want exercise stress cardiac MRI in order to see ischemia under true physiological conditions. But this is a very large topic on its own, and I won't cover it for today. So for obstructive CAD, vasodilator stress CMR is the go-to protocol. We will then discuss whether we should include rest perfusion or not. Coronary microvascular disease would require quantitative stress CMR, which we were hearing about a little bit earlier, and I will discuss this briefly. In this regard, rest perfusion may be retained. For anomalous coronary arteries, the butamine stress cardiac MRI is the most appropriate. It most closely matches physiological stress, especially the potential compression of the malignant coronary artery traveling between the aorta and the main pulmonary artery. Right. So now that we have discussed the protocols, we're going to try and answer clinical questions, and let's look at how we can prepare to make sure we perform quick and efficient ischemia imaging. So let's begin with the SCMR protocol. So this is a guideline paper that I would highly recommend you look at, and it was published around about three years ago. I've included a QR code so you can see the document for yourself. It gives lots of very important and useful advice about stress CMR, along with advice on different drugs used in perfusion sequences. So this is the 30-minute guideline, which was actually co-authored by our previous speaker, Professor Markle, and is also very pertinent today, which he has showed a little bit earlier. Again, I've included the QR code. So now that we've covered the key guideline papers, let's continue our preparation work by looking at investing in the team. So in order to scan rapidly, technologist radiographers are critical. I don't think this can really be emphasized enough. We need to train and invest in our radiographers. In doing so, this will allow us to scan patients rapidly, and I'll show you some of the little tweaks that technologists and radiographers who are very familiar with stress cardiac CMR can do in order to find that time to scan rapidly. And furthermore, we also, of course, need to support our nursing staff because they're the ones that help manage the patients and prepare them for stress cardiac CMR. So in our institution, we actually do we use checklists to make the workflow a little bit easier and a little bit faster, and here are just some examples where we use checklists for looking for contraindications for drugs. You can do this for other aspects of the stress CMR exam. We also perform ECGs so that we can ensure patients do not have second-degree heart block prior to giving, say, for example, adenosine. So in our unit, we also, this is not very common, but we actually do pulse gating. So when we're preparing the patients, not just do we do the ECG gating, we also attach the pulse gating, and I'll show you the reason why. So this is a patient that's about to undergo a stress perfusion, and what you can see is that the baseline is a little bit wavy, but now the perfusion sequence has started, and the baseline is actually being affected by the radiofrequency pulses. So this is what it looks like just before, and this is what it looks like after. Fortunately, in this circumstance, the ECG gating is working, so you can see this by the white lines here, but in some situations, if the ECG gating starts going awry, that could actually be quite problematic when trying to interpret the stress CMI examination. You might have missing slices, or you'll have different cardiac phases in your perfusion sequences. So this is something just to pay attention to. The technologist or radiographer will actually be able to identify this, because at the planning stage, when they're planning the slices, they will be able to see whether the radiofrequency pulses are affecting the perfusion sequence. So this is, they would know this before you inject contrast. If you discover this after you've injected contrast, it's a little bit late, obviously. So another point to make is that, from our experience in Hong Kong, we actually switch the scanning positions for pulse gating if we were to utilize it. So this is the normal arrangement that you would normally acquire. Stress perfusion, you'll start the basal slice, followed by mid-ventricular slice, then the apical slice. But what we find is, actually, we switch the order around. And the reason why is, particularly in Chinese hearts, the apex is quite thin. So if you acquire it in the traditional position, the apical slice is so thin that, actually, if you want to do quantitative perfusion, it becomes more difficult. So we switch this so that the apical slice is in the systolic phase. And this makes it easier for quantitative perfusion, but also easier for you to look at when you're doing qualitative analysis. OK, so the third part of my talk, I'll cover drug choice, because this is actually a very important aspect, and it will actually have an impact on your time savings. And I've just illustrated here the five drugs that are recommended in the SCMR guideline. So what this table shows is that regadenosine is, by far and away, the fastest at achieving stress and requires only one venflon, unlike all the others. But it comes at a higher cost. Adenosine is the next fastest, followed by dipyridamol and ATP, which stands for adenosine triphosphate. So for those of you that may not be so familiar, adenosine triphosphate and adenosine are not the same. The butamine takes even longer and will also require two venflons. So now what I'll do is I'll go into a little bit more detail about the various drugs so that you can take this into consideration when you're trying to find time savings. So regadenosine really is very fast, because it's just a single injection of 400 micrograms, and you just wait one minute, and then you can initiate the scan. You don't need to worry about whether you've achieved adequate stress, unlike some of the other agents. There are some downsides. Normally, most centers will give 100 milligrams of aminophilin in order to reverse the stress. There is also another issue that's becoming more apparent, which is even after 100 milligrams of IV aminophilin, within the time frame, especially when you're trying to scan very rapidly, the patient may not have actually achieved a resting state. And we kind of know this because I'm part of an international multi-center consortium called ACWA. And we basically have looked at sites with using ragged adenosine and comparing the quantitative values of the rest perfusion images versus adenosine. And we can actually see that there is actually higher resting perfusion myocardial blood flow in those that have ragged adenosine versus those that have adenosine. So this is one issue just to bear in mind. It also causes a stronger response. So you might have to be a little bit more careful with asthma and heart failure patients. But overall, this is going to be the drug that will save you the most time. So now I'm going to talk about probably one of the most popular drugs for stress CMR, which is adenosine. It has a very short half-life of seven seconds. It's cheaper than ragged adenosine. But although that seems like a good thing, the upfront cost is low, there's actually some hidden costs that you may not be aware of. One is because you have to check whether you get adequate stress. So if you don't get adequate stress, you'll have to increase the infusion rate. So this obviously lengthens your scan. There are also other costs, such as, for example, pairing the injector. The nurses have to do an extra venflon. And sometimes some patients are more difficult to cannulate than others. All these extra work that has to be placed on the team will obviously result in loss of time. So although adenosine might be cheaper in terms of an upfront cost, actually, overall, there may actually be either equivalents or may actually favor ragged adenosine for that reason, even though it's a little bit more expensive. So I'm aware of one site in the world that actually manages to do a single venflon. But there is actually one caveat to this. You just have to be a little bit careful that when you do a single venflon, sometimes, depending on which drug is preferred, basically, the contrast versus the adenosine, because they're both competing for the same venflon, you might, for a small moment in time, have more adenosine on board or just purely contrast with no adenosine. So that's just something to bear in mind. But that particular center that does this has a vast amount of experience. And it seems to work very well. So I'm going to talk about adenosine triphosphate because I know that, actually, in my part of the world, in Asia, this is actually a very popular drug. The reason why is it's cheap and it's easily available. However, again, this also comes with hidden costs. It takes longer than adenosines. And it's more frequent that you're going to have to increase the infusion rate in order to achieve adequate stress. So actually, for a rapid stress exam for my Asian colleagues, I'd probably try to persuade you that you should probably shift to one of the other two drugs if possible, if you want to go more rapidly. So as mentioned earlier, in patients with anomalous coronary arteries, dibutamine is likely the appropriate drug. If this is the situation, then one aspect that we found helpful to speed up the examination is actually to switch the order of the perfusion sequences. And what I mean by that is, essentially, we put the dibutamine at the end of the study so that, essentially, when you get to the stress portion, you can take your time to adequately stress the patient and then let the patient exit the scanner and recover outside rather than recovering in the scanner, which occupies more time. OK, so let's move to scanning and rapid protocols. So this is actually a very interesting abstract that came out of the Bart's Heart Center in London. And they have a huge amount of experience with cardiac MR. They do around about 10,000 a year. And what was quite interesting in this abstract that was published is you can see that, conventionally, on their conventional sequence, they already scan stress EMR cases in 36 minutes. Non-contrast EMRs in 15 minutes. This is just normal practice for them. So they are already scanning very rapidly. And what's interesting is, for their rapid protocol, they can do stress EMR in 22 minutes. And I also want to emphasize, they use adenosine. They don't use regadenosine in these studies. And if you look at the abstract in more detail, you'll see that, actually, their fastest scan was 14 minutes, which, actually, I find quite, that's just very impressive. 14 minutes is exceedingly fast for a stress EMR. And how did they do this? Well, essentially, they dumped their black blood stack. And they also used a rest perfusion, or they dropped their rest perfusion sequence. So also moving to another group. So they are, obviously, big lovers of T1 and T2 mapping. And they wanted to keep this as part of their tailored approach. But even they were able to reduce their scanning time, such that stress EMR was within 33 minutes. So you've seen this slide already, presented by Professor Markle. But one thing to just note was, actually, the SEMR has actually dropped the rest perfusion sequence. So this is something that you can certainly consider. But it's really if you're very experienced at looking at stress perfusion sequences. So another aspect to also consider when you're scanning, and this is something that technologists need to be aware of in terms of saving time, is the higher the heart rate, the faster the scanner will require 2R acquisition. This results in increasing acquisition time and potentially more adverse reactions. So I'm just going to illustrate how this works. So you have a patient with 70 beats per minute. The machine can put in three slices in each R interval. But if you get to 100 beats per minute, what ends up happening is the third slice gets placed after the second heartbeat. And as you can see, for the same amount of data, you need a longer time to acquire, which is a bit unusual for MRI. Because normally, faster heart rate means faster acquisition. So one of the things that technologists will do is they can sometimes cut the number of phases. And that way, you save time or maintain the same amount of time that you're going to use for scanning. So for those of you that were probably wondering, oh, no, so I can't drop the rest perfusion sequence, well, don't worry. This is the topic that Professor Markle briefly talked about. So I'm going to show you two patients. These two have stress perfusion. And you can see that sometimes the bane of all our lives with stress perfusion is sometimes when you have the odd dark rims here and there. But if you have quantitative perfusion, it becomes really clear which patient is normal and which is abnormal. The one at the top has good perfusion. The one at the bottom doesn't. And the other thing that you may have wondered about is, how long does the computer take to contour all of this? Well, with artificial intelligence, it has a high accuracy of 0.93. And it takes one second to do over 100 images. So in conclusion, we've been through choosing the right protocol in terms of answering the clinical question. We've also talked about preparation and the emphasis on investing in your team, doing checklists, consider pulse gating. Third part is drugs. And ragadenozone is probably the favorite drug in this context. Scanning protocol, consider dropping rest perfusion. And lastly, quantitative perfusion is really going to change how we do stress CMR. So as you can see, stress CMR can be performed in less than 30 minutes. With that, I thank you very much for listening. OK, good afternoon, everybody. So for those of you not doing cardiac MR already, wouldn't you like to start doing it? And if you do start doing it, wouldn't you like to do it easily and simply? And it would be even better if you can incorporate flow. So I will attempt to demonstrate this to you, that you can all start doing cardiac MRI in a very simple form. And you've heard a lot of evidence from our prior speakers on that subject. So to get started, I asked Dali to draw a picture of an MRI scanner on the left and a bicycle on the right. And just full disclosure by me, this is my first attempt at using Dali. And I was impressed that it actually drew some reasonable objects. And you're probably wondering, what on earth is the connection between an MRI scanner and a bicycle? Well, hopefully at the end of this talk, I can demonstrate to you that MRI is as easy as riding a bike, if you consider riding a bike easy. So we've heard a little bit about the standard CMR protocol already. And I'm just illustrating it here again. I would like to emphasize for everyone, though, that there is a core cardiac MR program. And it really, our protocol, and it really is cine imaging, inject the contrast, do delayed enhanced imaging. And that basic protocol can address most of the indications that you're asked or questions that you're asked of in a cardiac MRI. Now, nowadays, we add in mapping, we calculate ECV, and we do flow imaging as well. And suddenly, the protocol becomes quite cumbersome. And in this example, this is about an hour long. I can tell you in practice, for us running that protocol, it's even longer than an hour. And that's really not sustainable in our modern stresses on our workflow, where we have to scan quickly. We've heard about the rapid CMR protocol. And this is based around accelerating the core techniques. So on the bottom, you see real-time motion-corrected cine imaging, which covers the heart pre-breathing from top to bottom in a matter of minutes. You can inject the contrast. And then you do your single-shot MoCo delayed enhanced imaging. And you can really get beautiful results. Do all of this in less than 30 minutes. And also, if you want, add in some of these other techniques and other shortcuts to accelerating the overall protocols by giving the contrast early, which some of you are familiar with already. And we've seen this publication already. And I think it's OK to repeat it. But the rapid CMR protocol has now been published in various iterations and forms. And particularly, when you add in stress perfusion, as my colleagues have demonstrated, you can do this protocol very quickly and very accurately. So what are the approaches we can use to integrating 4DFlow into our standard acquisition? Well, one approach is just to accelerate the 2DFlow acquisition, like we've done for real-time cine and for delayed enhanced imaging. Secondly, we can add an accelerated 4DFlow acquisition to the rapid 4DFlow technique. So just accelerate 4DFlow imaging. Or thirdly, is to use a single one-click acquisition for 4DFlow and morphology. And I'll talk about that at the end. What is the ideal one-click exam, the one-size-fits-most 4DFlow exam? Well, we want a technique that can be set up uniformly across multiple scanner platforms. And in reality, that can be challenging, because sometimes we have different software versions. We want minimal technologist modification, so eliminating that prospective constraint that Michael talked about, where you just have a simple one-click axial acquisition. We'd like it to be free breathing. We want the image quality to be excellent. And we want full coronal coverage of the vasculature and all the intercardiac structure. So that's asking a lot. And we need a lot of coverage, a lot of speed, and a lot of high quality in our images that we produce. So as I mentioned, one approach is to accelerate our 2DFlow technique. And this is really going to real-time flow imaging. This can be achieved now by using a KT sparse acceleration. There's no need for ECG gating or breath holding. And you can acquire a single slice in about five heartbeats with a temporal resolution of 48 milliseconds. And you can see from this acquisition alone, you get tracings that are pretty comparable to the conventional phase contrast technique. And this has just recently been published in the literature in the European Journal of Radiology just this year, where they applied this technique and compared it to a conventional phase contrast 2D imaging technique, again, with that temporal resolution of approximately 50 to 55 milliseconds, and saw excellent correlation between velocity and flow measured, in this case, across the aortic valve and the pulmonary valve. So the second approach I mentioned is really adding 4DFlow MRI to your standard protocol. So here's an example here where we apply it to our standard MRI protocol. This is a patient with a bicuspid aortic valve. We're measuring the aortic dimensions. We've done a time resolved MRI on the left with a minimal amount of contrast. We do standard 2D phase contrast in the middle, and then we do 4DFlow MRI. And this really produces beautiful images. We can measure the peak velocity. You can see the aortic regurgitation here on the right on the 4DFlow MRI. But the challenge, of course, is that the acquisitions are long. There's a lot of prospective setting up of scan orientations, and then there's a lot of post-processing afterwards. So we can accelerate the 4DFlow acquisition, and this is some work by Liliana Ma, a few years ago, where the 4DFlow acquisition was accelerated quite significantly to less than two minutes and showed excellent correlation between the standard 4DFlow and the accelerated 4DFlow when measuring the flow at different points in the thoracic aorta. I remember showing this a few years ago at this exact same meeting, and this was very much experimental at the time. But I can tell you, we use this fairly routinely now in all of our, at least, vascular and aortic valve acquisitions. So one way of integrating this is just combining a rapid aortic flow protocol with a non-contrast MRI. So here's the non-contrast MRI on the top with full chest coverage. You see the 4DFlow acquisition on the bottom, and then you can go back and retrospectively interrogate the valves, are they aorta at any point, to generate velocities and flows. And all of this can be done in less than 10 minutes. This is without any Cine imaging of the heart. This is just a basic vascular protocol. We can also do this with a contrast-enhanced MRI protocol. And here's an example of a dissecting aortic aneurysm or an aneurysmal dissection. If you look at the MRI in the middle and on the right, it's hard to get a sense of the size of the false lumen, but it's only when you look at the 4DFlow acquisition, and you see the really wide-open finestrum right at the top in the upper-descending thoracic aorta that you see rapid flow into the false lumen. And it also shows you in great detail why that lumen, that false lumen, is expanding over time because there's a lot of flow from the true lumen into it. You can go and measure the exact velocity across the fenestration and measure the flow. And these velocity and flow measurements have been shown to be biomarkers of future aneurysm expansion, and potentially marks a particular indication use for the 4DFlow MRI in this population. We've also used 4DFlow MRI in conjunction with contrast-enhanced MRI in congenital heart disease. So this was an incidental patent ductus arteriosus, which you can see on the left here on the MRI. When we do the 4DFlow MRI, you can see the patent ductus also morphologically, and also the flow extending from the aorta into the main pulmonary trunk with a dilated pulmonary trunk. And from that, you can measure the QPQS to decide how severe that shunt is across the patent ductus arteriosus. So this is really just examples of showing 4DFlow MRI in conjunction with some of our other standard techniques. But what we'd really like to do is measure morphology and flow all in the same acquisition and be able to calculate various quantitative parameters from that. So this is an example from Albert Cao at UCSD, where he uses 4DFlow MRI in conjunction with the CineMagnitude images from the same acquisition. So this is 3D spore gradient deco in approximately acquisition time of about 15 minutes to measure the flow overlaid on top of the morphologic imaging. And here you can see a fairly rapid flow across an ASD. This can be measured again to measure QPQS or measure the absolute flow across the shunt. And this is in a patient with a prosthetic aortic valve and a prosthetic mitral valve. We can use this technique for other parts of the anatomy. Again, since it's 3D, since it's dynamic, you can go in and interrogate the data in real time, watch it functioning and contracting, and then hone in on the area of abnormality. In this case, this is a mitral regurgitation. You can visualize the eccentric regurgitant jet and measure the severity of that regurgitant jet. Again, all with a 3D CineMorphologic 4DFlow MRI dataset. Another approach is to do 3D Cine imaging in conjunction with 4DFlow MRI of the heart. So this is an example from UCLA where they've developed this technique called MUSIC, which is multi-phase steady state imaging with contrast where they can generate Cine acquisitions of the beating heart in more of a vascular visualization, and then go in and reconstruct standard orientations, in this case, a Cine4 chamber acquisition. And then with a second acquisition of flow, you can see an example of flow across an ASD here with the defects seen pretty well in the Cine images in the interatrial septum. This has been published in J.C. Amir a number of years ago where they applied this technique mostly to imaging neonates with congenital heart disease as a patient with a hypoplastic ascending thoracic aorta, also with a VSD, and on the Cine image, you can see the flow across the VSD showing the flow acceleration and the flow across the defect in the upper part of the interventricular septum. But then moving on, really what we're looking for is a full 3D acquisition combined with flow. And this is what has been achieved by Matthias Stuber in the Lausanne SHU group. Michael showed an example of this already. This is a free-running 3D SSFP technique. It's self-gated, so no need for cardiac leads, no need for breath holding. Use compressed sensing for acceleration. And you can reconstruct data with cardiac resolution to generate standard Cine images of the heart. You can measure or look at images across different points of the respiratory cycle, or you can take a simple image and reconstruct the entire dataset to look at the vascular anatomy in great detail. So tremendous versatility. In fact, I would say this is analogous to a retrospectively gated cardiac CT where you can go back in and reconstruct the Cine data in any phase you want, except you're not using any contrast at all. So Dr. Markle, who spoke here, and this is all in less than 15 minutes acquisition. So in collaboration with the SHU group, Michael's group has also managed to combine this with flow imaging. And again, showing all of this time-resolved Cine imaging of the heart, which can be sliced and diced retrospectively in any orientation you want. You can then apply 40 flow images to the entire acquisition. So you can retrospectively, in addition to going back in and looking at the orientations, you can retrospectively go back in and look at the flow data and measure velocities and flow as well at the same time. So this is truly realizing the one-click single axial acquisition of the heart, incorporating Cine imaging and flow imaging. So finally, I'll just finish up on addressing briefly some of the post-processing constraints. So many of us have seen this slide probably where AI can be implemented into the radiology workflow. I've talked about the acquisition side of things. We can also apply AI on the scheduling, for example, automated protocoling, automated scheduling. But really what we want is intelligent processing of the data. So we're not spending only 15 minutes acquiring the data, but then a couple of hours processing the data afterwards. That would make no sense at all. So a lot of advances have been made by the group here at Northwestern, led by Michael Markle, where they've applied AI and deep learning to the pre-post-processing and post-processing of 40 flow data, such that a lot of this processing can now be performed in about less than two to three minutes, which is really quite staggering when you think only a few years ago, it was taking sometimes 30 minutes to an hour to calculate and measure all the data afterwards. So a lot of advance in that area, a lot of, I think, promise on the horizon as we look forward to what 40 flow MRI can do. So to finish up, 40 flow MRI does allow for comprehensive push-button 40 flow MRI. It is now feasible with a single axial acquisition. You can integrate automated intelligence scan setup and automatic quantification to really improve the overall protocol. So to finish up, this is the final result from Dali. I asked Dali to draw a picture of an MRI scanner riding a bicycle, and I don't think that went so well. That's why I had to put in the MRI. So I think we got a bit more work to do with generative AI, but I will argue to you that we are basically nearly there with 40 flow MRI. We now have a single one push easy acquisition for a full 40 flow MRI of the heart. Thank you for your attention.
Video Summary
The talk covers advancements in cardiac MRI, focusing on reducing scan time and improving efficiency in acquiring cardiac data, specifically aspiring to achieve comprehensive MRI exams in as short a time as possible, around 30 minutes, and potentially even 20 minutes. The traditional approach to cardiac MRI, which involves complex and lengthy protocols, can now be simplified and accelerated through intelligent protocol design, modern MRI systems, free-breathing techniques, motion correction, and the use of AI and automation.<br /><br />Key aspects of typical cardiac MRI include assessing cardiac function, perfusion under stress and rest, late gadolinium enhancement, and optional flow imaging. New methodologies and technologies allow significant reductions in scan time, such as using parallel imaging to reduce breath holds, leveraging AI for protocol automation, and adopting innovative free-running 3D techniques.<br /><br />Numerous enhancements like real-time cine imaging, accelerated 4D flow imaging, and automated post-processing are making MRI more accessible and efficient. Continuous improvements in AI-supported workflows and free-running 3D techniques pose promising futures for easier, more efficient cardiac MRI scans.
Keywords
cardiac MRI
scan time reduction
AI and automation
protocol design
free-breathing techniques
motion correction
parallel imaging
3D techniques
RSNA.org
|
RSNA EdCentral
|
CME Repository
|
CME Gateway
Copyright © 2025 Radiological Society of North America
Terms of Use
|
Privacy Policy
|
Cookie Policy
×
Please select your language
1
English