What I'm going to be discussing is really high-resolution MSKCT and how we perform it. And I think what you're going to find is that a lot of what I'm going to discuss is really just a reiteration of a lot of the basics that I think sometimes people forget. So what we're going to take a look at is how to change and modify the acquisition, how to change and modify the post-processing, how do we change modify visualization at the PACS workstation or whatever sort of workstation that you're utilizing, and then also I really want to bring into play the multi-planar reformation function, which we can do individually as radiologists. We clearly typically have our technologists perform at the workstation, but it's a tool that we have to our access that I think sometimes we underutilize. So the acquisition. It's boring. I understand it's boring. I get it. But it's really the crux of what it is that we work with. The data set that's acquired is all that you have. You can play with it a little bit, but you really have to pay attention to how that acquisition is acquired so that you can use that data to the best of your ability. So there are a lot of parameters. Clearly we have KVP, we have the mass, detector element size, collimation, pitch, focal spot, and sampling frequency. The two big ones clearly are still going to be KVP and the mass. Those are going to be the two big ones that we always have to pay attention to. So our routine for most of our MSK work, and again, I'm really speaking specifically about MSK work, is at 120 KVP. For our larger patients and the setting of hardware, we will go up to 140 KVP. Notice I'm talking KVP and not KEV, which we would talk about in terms of spectral imaging. When we do that increase, though, we are increasing by 0.3 times the dose that's imparted. However, for the most part, we are involving the periphery and less radiosensitive areas. So the dose is particularly not the greatest concern of me. What's of greater concern of me is that we get good imaging at the one time that we perform the exam. For our younger patients, especially the pelvis, which we do a lot for impingement studies, we'll do 100 KVP, noting that when we go from 120 to 100, we're going to decrease that dose by 0.5. On the mass, larger patients, we're clearly going to have to use larger mass. For those of you that don't sit at the console, and probably most people don't, what happens is that the technologist gets a sense when they scan and do their scout, is what is, how much noise is there that's involved? And the computer will generate how much mass is going to be acceptable. So we want to make sure also that we have dose modulation on so that we can decrease the dose through the parts that are thinner and increase the dose through the parts that are thicker. There are clearly times where that cannot be applied, but it's something I think you always want to be mindful of. How do we keep that dose down to a minimum? So particularly larger patients and setting a hardware, unfortunately dose modulation has to be removed. And the mass is going to increase in a linear fashion as compared to the multiplicity fashion that you would see with KVP. So old slide, but it still works perfectly fine. And I think if you take a look, if we go from 80 to 100, what do we increase the dose? Two times. 100 to 120.5, 120 to 140, we do it by a third. So it seems simplistic, but it's always important to keep these things in mind. And when we look at the mass, remember, this is linear. If you double the mass, you're going to double the dose. You're going to do a half of the mass, you do a half of the dose. A third, you wind up doing a third. So again, nothing that's so revolutionary here, but I think it's important for people to get reiterated that these are the things that are really helpful. So what happens when we don't get enough dose? So this was a study that we got from the outside and then we wound up repeating it. And I think as you look at these images, I think all of us would say as we look at these images, they're pretty horrible. We have no sense of what is going on with the osseous architecture in terms of this patient that's post-operative for fixation of an olecranon as well as a distal humeral fracture. Really very little sense of what's going on with the architecture around the hardware, what's going on with the joint itself. And so we're left with a situation where we just don't have enough photons and enough energy to get a great sense of what's going on with this patient. Now, we then subsequently did this study and you can get, just already, you get a much better sense of what is going on with the hardware. I can get a sense of the spatial architecture of the trabeculae. I can see the bridging across the joint space itself of the humeral ulnar joint. When I take a look at the radiocapitellar joint, I have that joint space remaining. I have a big osteophyte here. I can get a sense of what's going on with the fracture healing that I have healing going on. I can look around that hardware to see if there's osteolysis or resorption around the hardware. So again, very basic parameters, but so important in terms of generating good imaging for diagnostics. So just looking at the two side-by-side, hopefully you don't get nauseous. But you can just take a look at the lack of visualization here with much improved visualization when we use much greater or enhanced technique. But I would say appropriate technique in this situation. So just a little bit about nomenclature that I think particularly for some of the younger people gets a little bit lost is multi-slice, multi-row, multi-detector, multi-channel. What is going on? So it's really probably the best terms are multi-detector and multi-channel. So instead of having one channel on the z-axis, we have multiple channels. And then those multiple channels are layered in rows to give multiple detector elements. I know in our vendors each panel will have about 2,000, 2,000 to 3,000 detector elements. We have 50 to 60 panels to generate all of those detector elements. So in schematic format, what you can see is that if you had just one detector, this is in the z-direction, you get just this one detector. Multiple channels, you can see that there are multiple channels in the z-axis. Again, multiple channels in the z-axis and then each one has a number of rows. And that's basically how that is set up so that you have just a generic sense of what goes on with this. I think just a simple concept of it can be somewhat helpful. Now the architecture of these, this was all 64. Clearly if you go up to 128, 256, whatever it is that you want to utilize, there will be different architectures or arrays of these elements. The important part with MSK imaging is we always want to use the smallest detector element to get our greatest spatial resolution. So the thinnest slice you're going to be able to acquire is at that smallest detector element. We can still overlap the images, but truly you cannot get any image that's going to be less than the smallest detector element. What this does is it allows the capacity to identify smallest gaps between the objects and gives us the possibility to achieve the greatest spatial resolution. The vendor will always tell you, oh, you have a spatial resolution down to, you know, 280 microns. If you look at photon counting detectors, they're saying 150 microns, even maybe less than that. But you have to optimize how that is acquired in order to get to that area. If you don't optimize the acquisition, you will not get that level of resolution. So the pitch traditionally we thought of as the table movement per rotation over the slice thickness. Now more so it's really the collimation. You want to get a sense of the table motion per rotation of the detector element width or the beam collimation. So higher pitch, that's going to allow coverage of a lot of area quickly. However, it does generate artifact. So what we've done is and typically is stated to use is the smallest pitch possible. Now, if you deal with areas with motion peristalsis, this can be an issue. Thankfully, bones and joints don't tend to move that much. And in our patient population, we don't tend to deal with a lot of disabled people. So we have a little bit more time. So I'd argue for the most part, the smallest pitch possible is going to be really to your benefit. The collimation, we're also going to use the narrowest collimation. Our settings on our units have two different collimations. It is going to be limited somewhat by the tube heating. So we have to be attention to that. But if we are able to use that, what it does is it allows us to get less averaging. So we have sharper images. So the way I think about that is similar to a resolution field of view over matrix. If I can keep, if I can get that bigger matrix and really a smaller field of view or collimation, I wind up with better resolution. So it is set partly by what's going on with the tube heating and the processing, but really the narrower collimations. So lowest pitch, narrowest collimation. How about the focal spot? We try to use the smaller focal spot, better in-plane resolution, and a narrowing detector to spacing. So it allows better geometric efficiency and then the better spatial resolution. It is tied to the dose in terms of what's imparted to the anode. So that if we get to a higher MAS, the machine will go to the larger focal spot. So again, this is something that will be directly linked to the mass that when the technologist is performing the study, but you want to make sure that that smaller focal spot is utilized to the best of your ability. And we will set the mass as high as possible so that we're able to still use it without going over to the larger focal spot for the most part. Larger body parts, metal, large patients, clearly is going to be an issue. And the frequency, a lot of vendors will have a standard versus a high-definition mode in terms of how many rotation, or excuse me, how much acquisition is a performed 200 to 2,500 views per cell rotation as compared to the standard of about 8 to 900. The issue with this is that noise is going to be a detriment. So if you have big noise generators, which are big people, and a lot of hardware, this becomes difficult. But by and large, when I can, I will use our high-resolution scanning mode to make sure that I get as much information as I can per cell rotation to help generate better spatial resolution. So let's talk a little bit about the post-processing and the algorithm. I'm not particularly smart. I'm very happy about that. We have two things. I don't have brains. I don't have lungs. I don't have admin. I've got bone. Bone is real good. So bone, high spatial resolution for a bone algorithm, lower contrast resolution. Soft tissue, we try our best, higher contrast resolution, lower spatial resolution. That's always going to be our trade-off when we're looking at the algorithm, the kernel, whatever it is that you want to call that. Okay, but that's really all that we have to think about. So for our non-metal cases, I will always use a bone algorithm, differing levels of how high spatial resolution. Most vendors will have levels that are greater and a little bit less, but still within the realm of a sharper kernel. Metal cases is pretty interesting. If you look past in the literature, everybody said, well, you have to use a soft tissue algorithm and then a bone window. We'll get to the window in just a second. I would argue I see the bony trabeculae so much better still on a bone algorithm, and by widening out the window, I'm able to still see around the hardware and the bony trabeculae. So I tend to play back and forth. I'll still get that acquisition. I use a bone algorithm and a soft tissue algorithm, but I wouldn't necessarily just blankly throw away the bone algorithm in the setting that I have underlying hardware. So let's take a look at this case. This is somebody who's had a total ankle, as well as a posterior, as well as a subtalar arthrodesis. So a fair amount of hardware sitting here in this location. And when I scroll through here, we have the bone algorithm and the soft tissue algorithm. And I'm just going to let these play. I think you get a good sense of what the osseous architecture is around the hardware. There's clearly going to be some streak artifact, which we can mitigate with the windowing and leveling. But if I look on the soft tissue, I don't see the trabeculae quite as well. I can see around the implant pretty well, but I don't see the trabeculae quite as well. And for that reason, I frequently will toggle between these two types of acquisitions. So I'm looking through and I can see, okay, I have a sense of what's going on with the bone and then around the hardware. And then I come here, I get a better sense of the trabeculae, as well as around the hardware. So if I take a look on just the one image again, around this hardware, I get a better sense. But here I see it pretty crisply. I can see the trabeculae. I don't really see trabeculae all that well. This person's also in a cast, which is going to limit my spatial resolution and my ability to take a look to some degree. But I still think that it's important that I can look at both of these. And sometimes I'll change my mind. In terms of this, I see a little bit too much of the artifact and I get a little bit better sense of what's going on with the bone interface to the hardware. Everything that we want to do is going to be at that interface. The bone to the hardware. What is going on at that location? And I would reiterate that to you when you talk to folks about development of CT. That's always going to be the crux of the matter. That's going to be where does your osteotomy heal? Where is your fixation for a fracture heal? What's going on with the incorporation of the hardware into the bone? That is really the money location that we want to be able to interrogate to the best of our ability. So using the thinnest slice as possible, we can then make our reformations in both the sagittal and coronal planes. We want to make sure... I'm a little over. So I'm just going to wrap up a little bit. So just very quickly, if you take a look here, you can see that when we do these two types of pole sequence or these two types of acquisitions, although they're at slightly different thicknesses, really not too much differences in how they look. We like to use about two times the detector element. So it allows us to take a look at things well, but at the same time not generate too many images. This was just a case that we had that was a ceramic on ceramic that completely shattered the ceramic on ceramic where I can still take a good look at what's going on with the bone and at the ceramic interface. And then I can also take a look here in terms of the bone at the ceramic interface. And then just standalone images. Just one mention about the center and the width. We like to increase the centering so that we can get it not quite as bright. And the width, we'd like to widen out the width so as to make sure that we have a longer array of image... of potential grayscale imaging to take a look at. And just one last thing about the oblique reformations. The technologists will typically do this at the workstation and we'll get these oblique reformations. But you can do it just as well. Take your NPR setting and then form your oblique reformations so that you can get a sense of what's going on with this scaphoid fracture in this position. So you're not clearly dependent or solely dependent on the technologists. This is something that we can clearly do very easily at the multiplanar reformation. So just a couple things we didn't even get a chance to get into. Mars. Beware of the surrounding radiolucency. The dual energy. I have not seen something where the spatial resolution looks great to me at the 140. And photon counting, we've seen a lot about that. And I think it shows great promise in terms of what we're doing. So for the Mars, you can see that radiolucency around those locations. And then this is a dual energy at 140 without even any hardware. I just don't see the trabeculae that well. And that's really prevented me from utilizing that to a great degree. So sorry for going over just a little bit but wanted to get to those points. So let's switch gears from electron interactions to proton interactions. And so my goal is to keep this as practical as possible and really show how we use 3D pulse sequences and how 3D pulse sequences build the base for advanced visualization which can be a great communication tool. So just quickly, what's the difference, right? So for 2D MRI, we acquire slice by slice. For 3D, especially TSE, but also gradient echoes, we acquire cubes, right? So every time we excite, we excite the entire volume which is extremely powerful, gives us a ton of signal. And if you keep your voxel size isotropic, that follows the principle that Eric just showed, you can do MPRs in virtually any way. You can slice and dice. You can curve. You can unfold. And you can volume render these images. And I think there's a lot of development going on here. So the paradigm is to keep it isotropic for the voxel size. Now, of course, the slice thickness is thinner and we have improved spatial resolution. For example, here, a small subchondral fracture. If you compare a one-to-one with 2D and 3D, you can see you do get more detail and you see these small fractures better. Would you have missed it? I don't think you would have, but you see better detail. So how does practical multiplanar reformation work? I think reading 3Ds allows you to adjust planes to structures like Eric just showed on your PACS system. PACS systems are very advanced these days. And so it allows you to highlight or work on findings yourself. So for example, here, a small meniscus tear. As you know, with 2Ds, you always hope that you hit that money shot here on your axials. But with your 3Ds, you can adjust it and you have enough thin slices to find these small tears. Now, the surgeon in this case said, I wish you wouldn't have seen that. This is a six-year-old and I don't want to see that meniscus tear. And now I have to do something because mama is upset with me when I say this doesn't matter. How about oblique planar reformations? So sometimes in 2Ds, we're trying to angulate to the ACL. We're trying to guess prospectively. Now, with 3D, we no longer have to do that. We actually have data sets. We can retrospectively do it. Everything is there. And you can do double, triple oblique angulations and you can highlight your ACL. You can go for the intermedial or the posterolateral bundles and you can see how you can highlight that. So that is all there. And I think all vendors have excellent 3D pulse sequences. Siemens uses space. GE uses cube. They have redone it. It's excellent quality now. And Philips has Vista and all the other vendors have those as well. You can also do axials that always run with an oblique structure. For example, here we have an adult-acquired flat foot deformity, posterior tibial tendon dysfunction, complex tearing there. And as you know, the PTT is just never in your plane. Sometimes our surgeons say, can you do like triple angulated coronals for me? And I said, I don't think our physics allow for that, but we can do 3Ds and we can actually run the axials along. Here you can see the posterior tibial tendon and you can highlight in thin space spatial resolution that complex tear there. And so you're avoiding partial volume averaging, which helps you with that, okay? And this works usually when you acquire with high quality. You can do the sagittal coronal axial. You acquire in one plane. We usually use the sagittal plane. And then here on the left-hand side is an ACL tear. Here in the center is a posterior root tear. And these usually work in multi-planar ways. And if you acquire with good quality, then your multi-planar image quality is being retained. So that's an important factor there. I will briefly touch on some technical considerations, right? So as you know, we do volume excitations, brings a ton of signal, but it also takes time. On the other side, we have a lot of signal to burn and we can accelerate. So there has been a lot of development with acceleration. There's two-phase encoding directions and you can accelerate and shift signals to optimize your signal gain. And then you can get under five minutes with high spatial resolution data sets these days. There's also compressed sensing, uses a different case-based sampling. It's using a smarter way. It says, in the periphery, I don't need all these points. I can reconstruct this mathematically. What counts, for example, here is the center and you can even gain a five- or a six-fold acceleration out of it. And what you can do with the different varieties is, for example, here, if you use a compressed sensing space, this is one that we use, you can get extremely T2-weighted pulse sequences and then you can get out that magic angle that's always nasty in the perineus longest tendon, right? So here's a magic angle effect. Here, with these long TEs, you get exquisite T2-weighting and you get this magic angle out of it. Similar here, here's a medial plantar nerve neuropathy. Here, you always wonder, hmm, this could be 55 degrees to B0. Is that really just signal, is the signal artifact, artifactual, or is it actually a neuropathy? And if you have these long TEs, which you can afford with these 3D pulse sequences, then you get exquisite T2 weighting, and that was a neuropathy. Of course, there's no talk without AI, and of course, we're working on making these even faster. And what used to be five minutes may soon be 90 seconds. There's a little bit of iteration artifact in there, but AI algorithms get better and I think we will see huge advances in acquisition time here post your root tear. And of course, you make this diagnosis, there's differences in image quality, but I think they're going to dwindle down. Now what about curved planar reformations? I think this is an extremely exciting topic for 3D MRIs. So this is an isotropically acquired ankle, and the flexor hallucis longus tendon is one of these curvy tendons that you almost never get in plane. What you can do with these data sets, and that's my practical point here, this is done on a care stream pack, so this is packs can do it as well, spectra packs can do it. There's a lot of packs that can do that. You can unfold these tendons. So you can plot in a curved fashion your MPRs, and then the viewers will unfold it and show that tendon in one plane. So that's what's meant with curved planar reformation, and that's afforded by the isotropic acquisition. So as you can see, you dot here the flexor hallucis longus tendon, and then it unfolds here, and then you get a data set where you can scroll through that tendon back and forth. And I think that is really valuable, especially around the ankle, but also other tendons, possibly even nerves, right? And so this is this unfolded flexor hallucis longus tendon. When I showed this to our foot and ankle surgeons, they said they had never seen that, and there is actually a bill you can, there's a code you can bill for that, right? So this also works with the anterior tibial tendon, extensor digitorum, and here peroneus longus tendon, which makes maybe the craziest turns in the foot anyway. So how about 3D MRI visualization? I think this is an exciting topic. It will allow us, similar to CT, we have always been a little bit jealous on CT. They have like cinematic rendering, global illumination, all these fancy stuff. We want to use it for MRI as well, and we learned these are great communication tools. It's easy to talk to other physicians, but also to patients, right? So what can you do? You can use those 3D isotropic data sets, and you can apply here cinematic rendering, and here is a peripheral nerve sheath tumor in the tibial nerve, and then you can simulate in situ if used, like you say, okay, it's with a surgeon. If you expose this, this may be how it looks like. Here is another one, a nerve sheath tumor you can see here in the continuation with a fascicular-like pattern, and then you can make these plastic appearances here. Here's a hematoma between the fascial layers in the calf, and you can plastically model that. I will say this takes a little time, but they can be automized at NYU, working on all the segmentation algorithms for that. Here's another one. You can do a virtual inspection. This is a clearly swollen ankle in a young individual with open physis. You can see here there is something under the periosteum, so that's a subperiosteal abscess. For demonstration purposes, you can 3D render that and maybe get a better visualization tool to explain to the parents what's going on. You can apply that in the knee as well. Here's an ACL tear. Here's a tear of the patella tendon at the tibial tubercle, and here is a bucket handle tear where you can see here the displaced handle medially. So there's a lot going on, and I want to close out with something that we have been working on for a while is virtual arthroscopy. I was searching and think this could be a tool to actually teach or preemptively predict arthroscopy, and if you have good quality 3D data sets, you can fly through the knee here. Here is a posterior root tear, almost horizontal in orientation, and you can do virtual arthroscopies and possibly also use this as a consent tool and tell patients this is what we're going to find. This is what we're trying to do, and hopefully use this gainfully to make our patient care better eventually. Thank you very much. I want to put in one plug. We just published a 3D MRI issue with seminars in skeletal radiology. There's a lot of great content in there if you want to follow up on a more anatomical basis here. Thank you very much. Let's work all in our alt scores and follow each other on Twitter. Thank you very much. So I'm going to be talking about dual energy CT, a few practical tips and tricks for you guys. So obviously when you talk about dual energy CT for MSK applications, really the first use of it and probably the most common use of it remains gout, so we'll focus on this. And actually this use has increased over the last several years because there's been new criteria for diagnosis of gout that was published in 2015 that actually included dual energy CT as a criteria for gout. So we are seeing more of this and not less of this over time. But outside of gout, you know, there's several applications where dual energy CT is very helpful in the MSK realm, and we actually published this recent review in AJR that covers this in detail. And I'm just going to touch on a couple of these that are more popular, which is evaluating for bone marrow disorders as well as for metal artifact reduction. So in terms of dual energy CT for gout, now obviously the clinical standard and the most common way gout is diagnosed is still just presentation and maybe aspirating the joint looking for crystals. So when is dual energy really helpful, right? So it's essentially helpful, I think of it as a problem solver. So it's helpful in the setting when it's an atypical presentation, maybe confusing to the clinician. If the aspiration may be in a difficult location or not possible. And also remember that sometimes up to 25% of patients who have a joint aspirated where there's not crystals found can still have gout. So if there's still suspicion of gout despite a negative aspiration, dual energy CT can be helpful. And also looking at extra articular disease where it's not necessarily a joint, which we certainly know gout involves, and dual energy CT is very helpful in identifying those locations. So here's a good example of where dual energy CT can be very helpful. So this is a patient, 65-year-old female, presenting with flank pain, radiculopathy, had a history of endometrial cancer, no previous history of gout. And this was done at an outside hospital. They did a lumbar spine MR, and you can see there's this low signal, T1, T2 mass, intermediate stir signal. And this was concerning for them for a paraspinal malignancy. And they also performed a CT at the time. Again, you can see that mass is hyperdense on CT. There's some bony destruction involved. And they actually referred this patient to our center for a CT guided biopsy. So when we reviewed the images prior to the biopsy, we were suspicious that this could be gout, even though gout's not typically seen in the spine, and this person had no history of gout. So we recommended that they get a dual energy CT scan. And obviously, you can see here on the dual energy CT scan, all of these areas where that hyperdense mass was and the low signal mass on MR is all lighting up as green. And you can see that this is indeed gout. And we were able to cancel our biopsy. So this has actually been shown in a recent review where axial gout is actually underestimated. Over 78% of these cases are actually only found at the time of surgery or aspiration. And they actually make the point that using dual energy CT can avoid unnecessary procedures or surgery. So that was a good use of dual energy CT for gout. In terms of a quick talk on how we do dual energy CT for gout, the most important thing to remember is with any dual energy scan, you want to try to get the largest separation or spectral separations, your high energy and your low energy beam as far apart as possible. So in most scanners, in modern day scanners, this is typically your high energy beam being at about 140 or 150 kVP and your low energy at 80. Most vendors now actually have built-in protocols for gout that you can just kind of turn on and use. One thing to remember, especially we use Siemens scanners and you can actually have a tin filter applied and that's been shown to help in terms of reduction of artifact. In terms of how we post-process these images, one thing to remember is even though you're doing a dual energy scan, it's important to do a regular post-processing of just your standard CT and your multi-planar reconstructions. And then do a 2D overlay where we can identify and localize where the gout is. And then finally, we do a 3D overlay. And so I think we do this routinely for all of our dual energy scans. So how good are we in terms of diagnosing gout on dual energy CT? Well, there have been multiple studies done over the years and several meta-analyses and this is some results from a recent meta-analysis that shows we're pretty good. The sensitivities in the low 80%, specificity is higher in close to 90%. So where we're not good, and this is important to remember, is in the acute setting of gout. So in the acute setting of gout, our sensitivity drops to just about 50% with dual energy CT. And this is an example of that. Here's a patient who had recent toe swelling and redness that was presumed to be gout, treated with some steroids. And you can see the ultrasound here, so I'm showing some synovitis at the second MTP joint. Then presented back in the emergency room now with ankle pain, had elevated white count, elevated inflammatory markers, also elevated uric acid. And of course, the question here is, you know, what's going on here? Is this septic arthritis or is this gout? So we said, well, since there's this history of gout, let's do a dual energy scan. So we did this dual energy scan. And you can see here, unfortunately, we didn't identify any urate deposits. But we went ahead and did a joint aspiration and crystals were seen in the joint aspiration. So this is an example of a false negative dual energy CT scan in a person who was having an acute episode of gout. And actually, in these situations, this recent article actually highlights the importance of looking at the non-contrast CT because you can actually look for hyperdense deposits in the area where they're symptomatic. And this can increase your sensitivity from that 50% range to close to 80%. I just want to quickly touch on a couple of other indications here. One is for bone marrow evaluation, like I talked about. This can be helpful for both bone marrow edema, as well as bone marrow lesions. So when we look at bone marrow edema, here's an example where dual energy can be very helpful. So there's a person presenting to the ED, hip pain, you know, pain after a fall. You can see the radiographs and the standard CT, it's very difficult to identify the fracture. But then on the dual energy scan, you can see there's extensive bone marrow edema here in the green color corresponding to a non-displaced greater trochanteric fracture. And this has actually been shown in a recent radiographics review that dual energy CT can be very helpful, and especially in the emergency room setting, to detect these occult fractures, hip, spine, wrist fractures. And so that's a very kind of useful indication for dual energy CT. How good are we in terms of diagnosing bone marrow edema? Again, we're pretty good. We're in the mid 80% sensitivities, you know, close to 100%. But again, keep in mind that it's not as sensitive if the injury happened very recently. So this hyperacute, less than 24 hours, the sensitivity drops again. And also, again, the TIN filter is helpful to increase your sensitivity when you're performing these scans. Dual energy CT can also be helpful for detecting bone marrow lesions. This has been shown mostly in multiple myeloma with high sensitivity and specificity. Here's another case where it can be helpful in terms of identifying bone marrow lesions. This is a case we performed a CT guided biopsy. You can see a PET positive lesion that left iliac bone. Hard to see on the standard CT. But on the dual energy CT, we can kind of identify that lesion, and then we can use that to localize for needle placement. And this indeed was a successful biopsy that showed that this person did have a metastasis there. And this has actually been published recently showing the usefulness of dual energy CT for identifying occult lesions when you're doing CT guided biopsies. But remember, when you're doing this dual energy CT for bone marrow evaluation, it's not as good with patients with a lot of red marrow. So here's a case. This patient had a lot of red marrow. It's all lighting up as if they have bone marrow, extensive bone marrow edema, and ephemeral necks bilaterally. But you can see on the MR, this is all just red marrow. There's really no marrow edema. And also remember, with this virtual non-calcium technique, which is what's used to do bone marrow imaging, it can only be used in non-contrast scans. So any of the contrast enhanced scans, this would not be an applicable technique. Finally, just to end on metal artifact reduction, another common use in MSK for dual energy CT. Here, what we're really talking about is virtual monoenergetic images. And the idea here being is that we want to reduce beam hardening, which is a cause of metal artifact reduction. And if you use high energy virtual monoenergetic images, you can reduce the artifact, as you can see here. And this is very helpful when you're talking about smaller hardware and less attenuating metal. So here you can see a screw from a hip arthroplasty. You can see there's artifact on the standard image, completely gone on the dual energy CT. You can see artifact here in a titanium humeral stem that's reduced on the dual energy scan. But it's not helpful when you're imaging large amounts of hardware. Again, here you can see a hip arthroplasty and a shoulder arthroplasty, cobalt chromium components, a lot of artifact. But the dual energy CT doesn't seem to reduce them. In fact, it seems worse. And that's because the artifact here is not being caused by beam hardening. It's really being caused by photon starvation, which is not reduced by dual energy CT. So just to wrap up some take home points in terms of for gout, it's a great problem solver. Remember, it's less accurate in the acute setting. And don't forget to look at the non-contrast studies to help your sensitivity. In terms of bone marrow, occult fractures and marrow replacing lesions is going to be very useful. And again, remember, it's not as accurate in patients with a lot of red marrow. And finally, in terms of metal artifact reduction, we're using virtual monoenergetic images to reduce beam hardening, which can be helpful, but not so much in patients with large amounts of hardware. Thank you. Thank you so much. I'll be talking about clinical dynamics in MR neurography. So tips and tricks will be in orange. Look out for them. The dilemmas that we'll be addressing are whether the nerve is abnormal. What do I tell my surgeon about a nerve injury? And is there malignancy in this nerve? So start with, is this nerve abnormal? I start assessing a nerve first by looking at signal intensity, because it's most often present, that abnormal signal intensity. So here's a normal nerve. It's slightly brighter than the adjacent skeletal muscle on fluid-sensitive sequence and iso-intense to skeletal muscle on anatomic or fat-sensitive sequence. An abnormal nerve tends to be bright, can be as bright as a slow-flow vessel or perhaps fluid signal. But brightness alone can be misleading. It can be seen, especially in older patients, in 13% of normal, healthy, controlled older patients in the sciatic nerves. So tip number one, fascicular hyperintensity alone should only be interpreted as clearly pathologic in conjunction with the clinical context. So here we have a tibial nerve in the tarsal tunnel. Looks normal. Slightly distally, we see that the medial plantar nerve, similar to what Jan had shown, is bright. But there's no change in caliber here. So beware of this location. Magic angle can impact peripheral nerves in addition to tendons. And unfortunately, sometimes we cannot prospectively increase the TE. Or even on STIR images, this artifact can persist. So this is a common site for artifact. If this is the only finding, be cautious about reporting it as abnormal in the absence of a correct clinical context. It has also been reported in healthy control patients from this study in NYU, where 70% of healthy controls had slightly hyperintense medial plantar nerve. Similar findings can be seen in the brachial plexus. These are normal brachial plexus elements. But the T1 nerve root, as it courses from the neural frame into the intrascalene triangle, just tends to be a little bit brighter. So if this isolated, and that's not the clinical question, this is often due to artifact. Sometimes vessels can mimic fascicular hyperintensity. So don't report that as abnormal either. These will have branching patterns and often flow voids adjacent to them. Next course, this can be impacted by anatomic variations, but I don't know of any anatomic variations that predispose patients to neuropathy. Mostly we're looking for extrinsic compression. So here's an osteochondroma deviating a lateral femoral cutaneous nerve. I do find caliber and size to be the most useful feature. This is an ulnar nerve that's been transposed, and it's enlarged. There's a narrow waist where there is scar and distal enlargement. So that's a very useful feature for scar entrapment or entrapment in general. And a fascicular pattern or nerve, or this normal nerve architecture, is this bubbly appearance that we see in a normal nerve. And it's often visible in a large caliber nerve, such as a sciatic nerve, and may not be seen in a smaller nerve, such as ulnar nerve at the wrist. But when present, it's a reassuring feature. It can be distorted in the setting of typically tumor or trauma. So this is a young woman with sciatica. She had blood products in her enlarged, distorted sciatic nerve at the sciatic foramen. And because of that, prospectively, this was diagnosed as neuromind continuity. However, she had catamenial symptoms. She did undergo biopsy, and this was found to be, based on biopsy results, imaging, and clinical context, endometriosis of the sciatic nerve. So clinical context is key. Perineural fibrosis in larger nerves, there's often a cuff of fat around the nerve. And it can be distorted in the setting of trauma or prior intervention. So these are axial images at the level of the inguinal ligament distal to it. This is the femoral nerve. We see that it's surrounded by scar. This patient had vascular graft surgery, and it's encased within the scar. So again, that's a useful feature, and this is why we have to include some form of fat-sensitive or anatomic sequence when performing neurography. And continuity is useful typically in the setting of trauma. So this is a bright sciatic nerve in a young patient, but there are no changes in caliber, and it's continuous. So this was a low-grade peripheral nerve injury. In contradistinction, we see a sciatic nerve that's enlarged, focally discontinuous and distally enlarged in this patient that had avulsion of their hamstring tendons. So this was a high-grade peripheral nerve injury. And sciatic nerve injury can be present in up to 30% of patients that have had a hamstring tendon avulsion. So nerve hyperintensity, though most frequently detected, is sensitive but not specific. And alteration in caliber and architectural distortion can add specificity. Also be aware of a magic angle artifact, because prospectively sometimes we cannot alter the TE. Next, what do I tell my surgeon about nerve injury? Peripheral nerve injury can be broadly classified using the Seddon or Sunderland classifications. Both of these rely heavily on histologic findings that are not visible by neurography at this time. So we cannot see axons. We cannot see myelin sheath or endoneurium. We can see fascicle, perineurium, or discontinuity. So I tend to classify these as low-grade peripheral nerve injury, and this works between the ordering physician and myself. And these will recover spontaneously without operative management. And on imaging, they manifest as a bright nerve with preserved architecture that's continuous and may be focally enlarged. On the other hand, high-grade peripheral nerve injury, which is comprised of either a complete or near-complete transection, and these are classified as high-grade because they will not spontaneously recover without intervention, have architectural distortion, enlargement, and plus or minus discontinuity. So here's a young patient that sustained a wrestling injury, and that is an evulsed fibular head fracture from here, and we see this wavy discontinuous common peroneal nerve. So that's a high-grade peripheral nerve injury. When discontinuity is present, it's reasonable to diagnose. This is another patient that has a hyperextension-type injury. We see a bright common peroneal nerve here. It's enlarged, architectural distortion is present, there's perineural fibrosis, but it remains continuous. So this was a neuromind continuity, again, a high-grade peripheral nerve injury. But higher up in the proximal field of view, we never quite see a normal nerve. So looking at the very top of the field of view, we see that there's an additional neuroma here. So that's important, and this may be encountered in the clinical practice of most people who interpret knee MRIs. People that have knee dislocation or multiligamentous knee injury can have peripheral nerve injury, and this occurs at three different zones. And typically, we look in zone two at the joint line or at the fibular head, but make sure that you see some normal architecture at zone one. So in this patient population, I try and make sure I see normal nerve architecture at that zone one. And if I don't, I'll bring the patient back, because if they intervene only here and there's an additional injury here, that patient is not going to recover. So for trauma, I tell my patients about architectural distortion, enlargement, and discontinuity, and try and classify the nerve as having high-grade or low-grade peripheral nerve injury. Next, is there malignancy in this tumor? So sometimes we're lucky and we see a case like this. There's a teal sign, a target sign, this person read the book, and we call it a benign peripheral nerve sheath tumor. We're not worried about malignancy. Sometimes we see a case like this. There's a cystic-looking thing at the expected location of the tibial nerve, and this was intervened on before she presented to us. What was not noted was this abnormality here, and shortly after her surgery, it recovered. So this is a sagittal oblique that I'd made, and this is the proximal tibiofibular joint. And we see a little neck, and so these are sequential images, and we see a neck coming from the proximal tibiofibular joint. So this is an intraneural ganglion. It can be diagnosed definitively on imaging. It occurs due to a capsular rent and a ganglion dissecting along the articular branch. This is not a malignancy. We're not worried. This is a young patient, and there's denervation, first thing to notice, in the short head biceps femoris, and abnormality in the common peroneal nerve contribution of the sciatic nerve. So this is a tip I didn't write in orange, but if the short head biceps femoris is involved, then you're looking at a common peroneal nerve problem that's proximal to the fibular head or in the sciatic nerve in the thigh, not at the fibular head. There's no restricted diffusion enhancement. This was a perineuroma, and it can be diagnosed with pretty high diagnostic certainty if these particular features are present. However, if you have an older patient with multifocal disease enhancement and PET avidity, you have to worry. This is not going to be just routine neuropathy. This was lymphoma. And lymphoma can affect the peripheral nervous system. It can be challenging to diagnose on routine testing, but positive on PET and MRI. So in our practice, if a patient is older, has multifocal symptoms, has a history of cancer, we tend to do either DWI or PET. This is another patient with an enlarged lumbosacral plexus and left sciatic nerve with enhancement and restricted diffusion, and yet a remote history of prostate cancer. So this was prostate cancer recurrence. These are rare, but do present a clinical challenge. So if the patient is older, has a history of malignancy or METs, and there's no target sign, we worry, and we try and use additional tools to make sure we're not missing malignancy. So in summary, the tips and tricks are as following. If peripheral nerve hyperintensity alone should not be interpreted as definitively pathologic unless there's a right clinical setting, beware of magic angle. Use these features to diagnose high-grade peripheral nerve injury and look out for those zone one injuries higher up in the knee dislocation patients. And if there's a history of malignancy or you're dealing with an older person with multifocal complaints, use some adjunct techniques. Thank you. Well, thank you very much for the invitation to the organizers. My name is Roman Guggenberger. I'm the head of MSK imaging in the University Hospital of Zurich, and apparently the only European in this session, although not by origin. Come on. I'm sorry. I'm an American. But even more thankful for that. So yeah. And I'm going to talk about ultra-short echo time and zero echo time imaging and synthetic CT in MSK imaging. We're first going to touch upon UT and 0T technique, what it's about, more or less. Just talk a little bit about that. Then also 3D T1 weighted gradient echo sequences for bone depiction. And lastly, we're going to briefly talk about synthetic CT, what's there to come, and how this could serve us in our daily clinical routine. I'm going to give you some examples in the spine and in the jaw, and maybe also some other joints. So starting off with UT and 0T, you may pretty well be aware that on conventional MRI, on T2 weighted sequences, we do have a problem with certain tissues like bone, the meniscus, or maybe even scar tissue, that these tissues do not have sufficiently enough signal on the echo times that we apply. So if we would shorten the echo time to, let's say, one millisecond or even below, we would be in a range where we could still have enough signal from those tissues to measure. And this is why it's called ultra-short echo time or even zero echo time imaging. Let me show an example. This is like a sagittal cut through the lateral compartment of a knee. You see a PD weighted sequence, an echo time of 35. If we now shorten the echo time to about four milliseconds, you can see suddenly the meniscus brightens up a bit. We do see some spots in the meniscus that do have an intrinsic signal, but interestingly, we also see some structures in the cartilage that we would not see on the standard sequences. And if we even shorten the echo time even more, maybe to almost zero, you can see how much signal we get from the meniscus. It's almost bright, but there's still some spots remaining in the meniscus. And in this case, interestingly enough, these were calcifications or crystal deposits in the meniscus. And a colleague of mine in his fellowship in San Diego, he correlated those deposits to micro CT analysis. And this is quite an interesting application of UT sequences for depiction of crystal deposits in the meniscus. But we may not only see those crystals in the meniscus itself. We may also find those crystal deposits outside the joint. This is an example of a gout joint, as we've already discussed in the CT applications by Naveen. You can see in the first metatarsophalangeal joint, this gout topos. You can see those faint calcifications in the topos itself and also adjacent changes in the head of the first metatarsophalangeal. So this zero T sequence obviously is an adjunct to traditional sequences and could help us in really identifying faint calcifications and adjacent changes in the bone. But coming back to the spine, you may know that with traditional spin echo or fast spin echo sequences on a T1 sagittal, we often have a hard time telling what's going on in this interarticular portion, let's say at the L5 level. Is there like a spondylolisis going on? Is there any other problem there? And again, this is a nice cadaver study where members of my group took a cadaver spine and then they cut this portion through. And this is what it looks like on CT. And if you then were to perform UT imaging of that spine, you can see that cut really nicely with almost the same edges as you would see it on a standard CT. And if you invert the grayscale, you have like this pseudo CT appearance. And this could certainly add value to your standard MRs of the spine if there's a question of spondylolisis. This is another example from our daily routine. You can see at this L3 level, there is some problem there. We were not sure really what was going on in that dorsal aspect adjacent to the spinal canal. We thought, okay, there might be some bony defect there, but I've never seen a bony defect like this. So we sent the patient to additional CT. And this is quite an interesting thing. He had like a still intact bony bridge behind the spinal canal, but there the spinous process seemed to have detached from that bone ring. But the interesting thing here is if you may not be lucky to have a UT sequence at hand, you can just perform a 3D T1-weighted sequence. There it goes. And it's really astounding how well you can delineate the bone in those sequences. If you invert the contrast, like again here, you see that defect in that dorsal bone. And also on the sagittal, you would nicely see that there is a distance in this part of the bone. We've done studies comparing UT sequences with 3D T1-weighted gradient echo sequences from Philips. It's a special product that they call the fracture sequence. And we wanted to compare those sequences with regard to certain distances about the cranial cervical junction. And just to sum it up, all measurements did not differ substantially from each other. They were really comparable to the gold standard of a CT scan. Also signal-to-noise ratios, both the UT and the fracture sequence were quite comparable. However, the contrast-to-noise ratio seemed to be a little bit higher in the fracture sequence as compared to the UT sequence in this case. Going a little bit further down in the spine to the sacroiliac joints, this is quite a delicate region, as you may well know. We're often confronted with the dilemma, is this degenerative? Is this post-inflammatory? And we do know that CT really plays a role there in depicting those subchondral sclerotic changes or maybe picking up like those erosions. And also in this case, 0T sequences could help us in defining what's really going on there. So we are currently performing a study comparing 0T images to a 3D T1 gradient echo. We call it the black bone sequence. Both are pretty comparable in the acquisition time, maybe the black bone a little bit faster than the 0T. And these are just some examples. You can see subtle sclerotic changes in the iliac bone on the right side here on CT. You can nicely spot that on the 3D T1 black bone and also on the 0T. You can also spot those bridging osteophytes here on the black bone and also on the 0T, although with a little bit of hazy signal there. And inflammatory changes like this erosion in the left iliac bone, you can nicely see it on the CT and on the black bone and maybe also on the 0T. But what is also interesting is if you have small air inclusions, as you may occasionally encounter in degenerative changes, those air inclusions, they also remain dark on the 3D T1 black bone or on a 0T sequence. And if you invert it, it just looks like a bony island. But in this case, it's just an air inclusion. One topic that we've really been investigating a bit more in detail is medication-related osteonecrosis of the jaw. May not be really something that MSK people have to deal in their daily practice, but we just came about a bunch of patients in our practice, which is why we got interested in that topic. It's actually a necrotic area of the jaw, either in the mandible or in the maxilla, that is there as a result of patients on antiresorptive treatment, so either by phosphonates or adenosumab. Those patients are really at risk of developing those necrotic foci in the jaw. And what is interesting, these foci are usually associated with quite marked changes of the adjacent mandible or the adjacent maxilla, but usually they're found in the mandible. As you can see here, increased sclerotic changes of the spongy bone of the mandible, increased periosteal thickness, and in advanced stages of those necrotic changes, you can of course find osteolysis or disintegration of the bone. So all these changes you can nicely pick up with UT sequences, and we were interested if those imaging findings would somehow compare to the reference standard. Like head and neck people, these days they perform CBCT a lot, cone beam CT, and we co-registered volumes of patients who underwent CBCT and UT imaging, and nicely enough we could see strong correlations of those gray values and signal intensities from both modalities to each other. But again, as I pointed out priorly, be aware of some small air inclusions, which may not be a problem here in the oral cavity, but if you really have a necrosis with adjacent air inclusions, these may cause artifacts and you may think of a bony fragment while it's some small air inclusion instead. You can even try to accelerate your UT sequence there by reducing the number of radial acquisitions, but you can see how substantially you can decrease the scan time from 315 to about 65 seconds, while still remaining the main features that you would find in such a jaw like disintegration of the bone, periosteal thickening, and increased bony sclerosis. And also artifacts may somehow come down if you decrease the number of radial acquisitions. So last few slides on synthetic CT. I have to admit we don't use that in our clinic. I had to do some literature search on that, but there's one term that you often come about when reading about synthetic CT, and that's cyclic generative adversarial neural networks. I'm sorry for that, but the abbreviation would be a CGON, and a CGON is actually a pretty cool tool. It's more like a convolutional neural network that really tries to generate CT images out of a learning set from MR images, and it typically consists of two parts. So we have the generator part on the one hand, which really tries to learn from those images to generate a CT image, but we also have a discriminator part in those constructions that then look if that synthetic artificial CT really matches the ground truth CT, and it would then feed back into the generator that kind of information, so that in the end you really get the optimum out of it. You can feed in such a CGON different kinds of MR images, be it T1, be it T2, anything that you want really your CGON to work on, but the ultimate goal would be to really come out with a synthetic CT image that looks just like a true CT image. And there's some nice studies out there using that technique, and again, I picked that one of the sacroiliac joints because I really found it interesting. As I said, post-inflammatory changes, we primarily look at them on T1-weighted sequences. We really are interested in that fatty metaplasia, but on the bony side in that subchondral sclerosis and probably even identifying some kind of erosions. We know that true CT can help us there, but this is what the CGON came out after performing their convolutional neural network. There were really awesome images that they've shown in their publications devoid of any signal noise, and I think the anatomy really looks 100% like it does on CT, and they've also done a diagnostic performance study where they found that this synthetic CT really helps in the diagnostic performance for the detection of structural lesions in the sacroiliac joints if you combine it with the routine T1-weighted images that most of you have in their standard imaging protocol. So summarizing it all up, we've talked about UT and 0T, what it is about, why we perform those sequences. If you don't have access to those sequences, you might use a fast 3D T1 gradient echo sequence. It can really deliver nice contrast on the bone, especially if you invert the gray scale and then have like pseudo-CT images. And lastly, we've briefly touched upon synthetic CT, which are now coming up ever more, and that may soon enter our daily workflow from some of those providers out there, and you've seen some examples in the knee, the feet, the spine, and the jaw. Thank you very much.

advanced imaging

high-resolution MSKCT

MRI diagnostics

image acquisitions

multi-detector CT

3D pulse sequences

dual energy CT

artifact reduction

synthetic CT