Hi, I'm James Galt, and today I will be talking to you about instrumentation in cardiovascular PET. Here you can see my disclosure. I receive royalties from the sale of Emory Quantum Software for analysis of MAG3 rheumography, which has very little to do with today's topic. Our learning objectives at the end of this presentation, you should be able to describe the components of a PET scanner and what they do, explain the differences between dedicated PET and hybrid PET-CT, and identify the quality control requirements needed for high-quality PET. The topics we'll cover are coincidence imaging, PET scanner design, differences between dedicated PET and PET-CT, we'll briefly touch on advanced scanning techniques, and we'll end up with quality control. While most people are familiar with mechanical collimation used in nuclear medicine and SPECT, where we have a collimator that is basically a lead sheet with parallel holes in it, and photons that are emitted from within the body pass through those holes. When they reach the crystal, a scintillation event occurs, and if the photons are not parallel to those holes, they're eliminated by the collimator. In fact, 99.9% of our photons are eliminated by the collimator. In PET, we do away with the collimator, which helps us with sensitivity, and we do a little better detecting photons, because we do not have that lead collimator, but we depend on electrical collimation or coincidence imaging to build our images. So we have the dual photons at 511 KV that are emitted after a positron-electron annihilation, and those two photons, if they're picked up by two different detectors, then we make the assumption that the coincidence event or the annihilation occurred in between our two detections. Now for both modalities, we have the problems of attenuation, where we lose photons, and we have problems with scatter that causes us to misposition photons. This is the way coincidence detection works. We have our positron-electron annihilation that produces two 511 KV photons at about 180 degrees. If we detect both of them, as shown in this top diagram, the difference in time will be very small. They're said to be in coincidence, and they will be counted. To determine if they're in coincidence, we set up a coincidence window of 6 to 12 nanoseconds, which is really based on the speed of light between the two detectors. And if they're both within that, then we say they're a coincidence. If we have two different coincidences that get picked up, it is very likely that in time, the difference between the photon at C and the photon at D will be outside of our coincidence window. We only pick up one photon, and that is called a single, a single photon counted within the window. If, however, they happened at so close to the same time we pick up both of them in that coincidence window, it's called a random, two unrelated photons counted within the coincidence window. And as you can think of the probabilistic nature of radioactive decay, the higher the count rate, the more likely it is that we're going to have randoms in our detection. Well, things that degrade PET, we've mentioned attenuation, but compared to SPECT, we only have to lose one photon before the pair of photons is lost, and we don't register a coincidence of that. We have Compton scatter photons interact with a outer shell electron. They change direction, they lose energy. If they lose enough energy, then this is part of our attenuation, it is not counted. But most of the time in one single Compton scatter event, we don't lose enough energy to be excluded, and now we'll connect the dots between the two detectors, and you'll see we'll misposition that coincidence. And in randoms, we have those two unrelated photons, and in the case of randoms, we might even connect the dots between where the two photons were detected outside of the patient altogether. We'll move on to PET scanner design, talk about PET detectors, and then about the overall scanner design. Here you can see a PET detector module. We have the scintillation crystal block right here, and that can be either an array of individual crystals or a single crystal partially scribed into individual crystals, such as the one that is seen here and the one in my diagram. Behind that, we have photomultiplier tubes. They detect the scintillation events in any crystal, and the crystal where the scintillation event occurred is identified based on the relative light distribution between these photomultiplier tubes. Very similar into the way that positioning is done in a gamma camera, except now we're only identifying the crystal. We don't do any positioning within that crystal. There are two types of crystals in use in PET today, BGO and LSO or LYSO, which are essentially the same thing. For sodium iodide, the crystal shown here at the top, that's what is used in a standard scintillation camera, and it's here for comparison. Stopping power you can think of is how many photons are we going to stop, and in one centimeter, which is a fairly thick crystal, we're going to stop about a third of the photons at 511 KUV with sodium iodide. That's why we're not using it for PET. We have a decay time, that could be considered the amount of time it takes for the flash to occur for scintillation, and the relative light output. For comparison, it's 100%. BGO, for many years, was the tried and true crystal for PET scanners. It has a very high stopping power, so we stop nearly all of the photons in just one centimeter of material, but we have a decay time that's longer than sodium iodide and very poor light output compared to sodium iodide. LSO-LYSO, on the other hand, has a stopping power almost as good as BGO, has a much shorter decay time, and that's very important later on in this talk when I begin speaking about time-of-flight imaging, and have a relative light output much higher than BGO, not quite as good as sodium iodide, but that's our crystal with a dense, fast, and a high light output. Well, once we have those detector modules, here you can see an array from one manufacturer of eight detector modules that are mounted together in one mounting so that they can be placed into a ring and form a ring in the scanner, and you notice that there are four across, and this might be considered a four-ring scanner. Over here on the diagram, just a few of the many, many different ray paths that might be detected in coincidence using this ring of detectors. Well, how do we build an image? We start by building a cyanogram, and a cyanogram is used to create a transaxial slice. A cyanogram is an image of coincidence events for PET imaging, and it gives us angle versus radial distance. So if you look at this graph right here, angle versus distance from the center, and in the cyanogram that you see here of a myocardial perfusion PET scan, distance from the center here, angle this way, and here you can see the heart scribing out a sine wave in that cyanogram. The cross-hatching here is the gap between the detectors. Well, this is how we build that cyanogram. In our first instance, we have coincidences directly across from each other, straight up and down, which we're going to call zero degrees for the purpose of this discussion, and a certain distance from the center, and I will plot that in the cyanogram that you see here. Then B, a little closer to the center, but an angle of about 30 degrees, and I will plot it here. Our next one is at a 90-degree angle, further away than B from the center, I plot that here. D at about 120 degrees, plotted here, E at about 150 degrees, plotted here. Now after getting tens or hundreds of thousands of these coincidence events, then that builds the image that you see right here, and as I said, this is the information we need to create one transaxial slice for our PET scan. Well, next, there are two different types of PET scanners in use. 2D scanners are not being sold anymore, but there are a great number of them still in use today, and so these scanners have SEPTA, and the purpose of the SEPTA is caused by the type of acquisition we're doing in 2D, where we only allow coincidence events to be straight across in the transaxial plane, so a single detector ring. So we put these lead SEPTA in there to screen out off-angle photons, so we have less photons that would be singles or would not contribute to our image detected, and that's the purpose. They just screen out off-angle photons. Well, in a 3D scanner, there were scanners sold for many years that could be 2D or 3D, and those would allow you to retract the SEPTA. Most scanners today are dedicated 3D and have no SEPTA, but we allow coincidence events to happen across planes in the detector rings. So in this, that increases our sensitivity greatly, and the only place we have lead shields is at the edge of the field of view, and the size of those might be tempered to increase the bore of the camera, but there are shields that would, say, screen out photons coming from the patient's bladder if you're doing cardiac imaging. Well, characteristics of 2D scanners. We restrict our coincidence to a single ring, so straight across, and what that means is that no matter where I am in this detector, I'm restricted to coincidences that are straight across, and I have uniform sensitivity across the scanner, so my count rate would be the same if I was imaging not a patient but a uniform phantom. Compared to 3D imaging, which I'll talk about next, they have fewer counts, but that might require you to use a higher pharmaceutical dose than a 3D scanner to get image quality that matches. Compared to 3D, we also have fewer problems with scatter and fewer problems with randoms, and that is why, in my experience on older scanners, I seemed always to be better off doing 2D imaging rather than 3D. 3D scanners, on the other hand, allow coincidence events across plane. So if I'm here in the center, I have a great amount of sensitivity because I have a fairly high acceptance angles where I might have coincidence, but when I get out here close to the edge, I have a very narrow angle, that means my sensitivity near the edge is much lower than my sensitivity in the middle, and as I say it's much lower near the edge, realize that it's comparable to that of a 2D scanner, so my overall sensitivity is much higher, and that means I also have a higher scatter fraction and more random events may be detected. Now this slide shows what that means in the real world. Our sensitivity is greatest in the center of the camera and the lowest and superior and inferior slices, so here I show the four most superior slices, the four most inferior slices, and four slices out of the middle, and since I have lower slices at the edge, I have fewer counts and that means more noise, and what I mean by noise in a digital image is the speckle that I get here, the degraded image quality. Now one thing that's important about this, if you're doing whole body PET, which even in the cardiac arena you may be called on to do as time goes by for something like sarcoid imaging and new imaging techniques to come, we often have to overlap the adjacent bed positions in order to compensate for that lower sensitivity at the edges, and that's a bed position as the patient moves through the gantry. They don't move a full length of a detector, but maybe three quarters of the length of the detector so that there's some overlap. Our PET scan resolution, well first, and something we can do very little about is positron range. Our coincidence detection provides the location of the beta plus annihilation, not the site of the beta decay, and the squiggly line shows that. The higher the energy, our beta plus, the further it's likely to travel before it annihilates, and for some types of imaging it might travel two or three millimeters, maybe even four to five millimeters, but that means we're now not imaging where the decay occurred, that's where a radiopharmaceutical is, but we're decaying where it annihilated. So that's a basic limit on our resolution. The next thing is our electron and our positron are moving when they meet, we have to conserve momentum, and that gives us a little bit of deviation from the 180 degrees that would be ideal for our 511kv photons. And so we detect those, we draw the dotted line, and we misposition by a little bit. Now if you look at the geometry, you can see if I make this ring larger, that deviation will increase due to that non-collinearity. And lastly, we don't do any positioning within an individual crystal. Crystals in use today may be three to seven millimeters on a side, and that is another limitation on our resolution. So let's talk about scanners themselves, dedicated PET and PET-CT, highlighting attenuation, the difference that we correct for attenuation in these two different types of scanners. Well first I want to talk a little bit about what we mean by attenuation in PET and how it differs from SPECT that people are maybe a little more familiar with. And here in SPECT, I have one photon that's emitted, and I'm only worried about one photon. So over a 180-degree orbit, you can see that I have a set distance, depending on where the camera is, that the photon has to travel. I have different materials. But regardless, I have a fairly short attenuation path length. In PET, I've got to get two photons out of the body. One going this way, which is very close to the surface of the body, but then that means the other one has to travel a long way. And our attenuation path length then is the full distance along that angle through the body. And it doesn't really matter where it is, since I have to get both photons out, the attenuation along that angle is the same. In fact, I could even put my event here outside of the body, and I would still have the same attenuation along that ray path. Well, that makes attenuation easier to correct, because it doesn't vary along that path, and I can actually measure it. You could think of it, if you think of mathematical equations and the solution to the image reconstruction problem, I have one less variable in PET than SPECT, and that makes a great deal of difference when I'm trying to correct for attenuation. So two ways of doing that. One thing I note, if I'm doing cardiac PET, the attenuation is just too great for interpretation of the non-attenuation corrected images. That means I have to know the distribution of the body, and that means the transmission scan. Dedicated PET systems use a rotating rod for that map, and PET-CT, of course, uses the PET-CT. Now, the rotating rod system built into the gantry is a rod of radioactive material, a field source. It moves around the patient, and that performs the transmission scan. I actually am measuring the attenuation at each angle. On PET-CT, on the other hand, I have a CT scanner in tandem with the PET, and I cannot simultaneously acquire the transmission scan and the PET scan, and that can be difficult to do with the dedicated PET as well. Not only that, but I have to move the table, and if you do a lot of patient scanning and you're there, you notice every time the patient moves, sometimes patients take that as a clue that it's time that they can wiggle. Here you can see non-corrected and corrected PET scans, and the way to tell if you're looking at a non-attenuation corrected PET scan is the lungs actually appear bright because the lung tissue is much less dense than the surrounding tissue, so it shows up as brighter than the surrounding tissue and uncorrected, and some people call this the bright skin effect where the outside of the body forms a rim. So comparing the two, well, dedicated PET systems don't have a CT scanner. They have that seal source, so they're a smaller footprint. They're less expensive. They have longer transmission times, can be up to five minutes, and easier attenuation correction because we're actually measuring attenuation. Hybrid PET CT, on the other hand, a CT for the transmission scan, tandem units, larger footprint, CT scanners are expensive. The better the scanner, the more expensive it is. Very short transmission times and more difficult attenuation correction for two reasons. One is done quickly, two, the energy is different. Well, for seal source transmission scanning, there are two methods there. Germanium-68 decays to gallium-68, positron decay, fairly long-lived half-life for germanium. And the energy, that's the attenuation, is the same as the annihilation photons. So we're measuring that attenuation, but our activity is limited by the detectors that are close to that source. So we have a very real limit on how much activity we can use here. And that means that our counts for transmission scan are limited as well. Cesium-137 is an attempt to overcome that, 667 KUV gamma ray, but not a positron emitter. Each detected photon, though, can be thought of as in coincidence with the line source. The energy is close enough to 511 that we have an easy transformation of attenuation coefficients. And now we're only limited by the count rate and the detectors across from that line source, not the ones that are close to it. So we can use a higher activity line source for the transmission scan. Here are some examples of a seal source transmission scan. This is the germanium-68 gallium-68 rotating rod provided by James Case in the publication in the Journal of Nuclear Cardiology in 2017. Here is a patient with normal body habitus. This is a morbidly obese patient. You can see the quality of the images. These are CT images. They're much poorer, you might think, than a CT image that you would use for diagnosis, but that's not the point. We're using this for attenuation correction, and the resolution is about the same as that of our PET scan, but you do have to be careful when you get into larger patients with the quality of the image that you get and any truncation that you may get from that system. Next, PET-CT. Here you can see the PET scanner, the PET detector modules, but what you can't see over here to the right is the CT scanner that's shown here. And here you can see the x-ray tube. Here is the x-ray detector on this side, and once the covers are on, it looks like one big unit, but PET is back here, CT is up front, and they're pushed together with one gantry. And here you can see the CT with the high resolution here, the non-corrected PET scan. So my trick is now to use this scan to attenuation correct the PET. And I have to be aware of several things, one of which is the difference in geometry. For SPECT, I have one photon, I have my gamma camera with the collimator, as we've discussed. For PET, I have two photons that are emitted, I have to detect both of those. For CT, I have an x-ray source on one side of the body, and then I have a detector on the other side and a fan beam geometry, and this rotates around the body. In PET, I have a ring of detectors, and for SPECT, my scanner rotates around the body. But with CT, I'm not measuring photons emitted for it in the body, I'm basically measuring the shadow that I see on this detector of the material in between. Geometry is not the only difference. I have an energy difference between the modalities. If I want to start here to the right with PET, I have 511 kV annihilation photons. I have Compton scattered photons from in the patient, and some in the detector. For technetium 140 kV photons, I have the spectrum here, again with Compton scatter, and the x-rays create this little peak there. But for CT, I have an x-ray tube voltage that would be somewhere between 80 and 140 kEV, and regardless, even if it's 140, my average energy is going to be much less than that kVP that I set at my scanner. That represents the maximum energy. I have much lower energies than PET and even SPECT. I have to convert Hounsfield units from CT to attenuation coefficients, and realizing that my average CT photon energy is less than the gamma energy for nuclear, much less than the 511 kV PET photons. In low-density materials, it's not that big a difference. It's all attenuation is due to Compton scatter. And higher-density materials, I start having photoelectric effect in the CT scan because of the lower energy. PET is still Compton scatter, but that means that I can't use a simple linear relationship to convert Hounsfield units to PET attenuation coefficients. I have to do a bilinear relationship with the change in the curve right there at zero, which is CT number for water. So above water, I use one line. Below water, I use a different line. This slide shows non-attenuation corrected images on the top row, attenuation corrected on the next row. This for rubidium myocardial perfusion scan of a normal male. And here you can see the overlay of the attenuation corrected PET on the CT. And what I want you to note is in the non-corrected one, how much lower accounts the inferior wall of the heart is than the anterior wall. Same thing for rest as shown here. So that's why we don't look at the non-attenuation corrected PET images when we're doing myocardial perfusion imaging. And we only look at the AC images, such as the one shown here on the right. This slide illustrates the motion of the heart and the rest of the body during respiration. That's from an FDG PET scan for an oncology study. But it shows you that the heart moves not just up and down with the diaphragm, but is actually moving in all three directions during respiration. And that's something that's very hard to deal with and to correct for. But methodology is coming for that. Before we get to correction for respiration though, we have to deal with the non-corrected or non-respiration corrected scan. What just happens when you acquire a standard PET CT scan? Well, the CT images are acquired in seconds. The PET images take minutes to acquire. And here I have a full inspiration breath, whole CT on the right. On the left, the patient was told to breathe shallowly during the CT and during the PET. And you can see increased volume in the lungs with the inspiration. Here in the sagittal, I see that increased volume. I also see that the diaphragm is very depressed compared to the shallow breathing scan. And the yellow arrows show that the spine is basically in the same place in both scans. Over here on the right, I can see also how now the lateral wall of the myocardium is out in the lung space. It is not in CT space. And that's a problem. So the asthmic recommendations for breathing are indexpiration breath hold or shallow breathing during the scan and that software alignment be available to correct for the remaining misalignment because there will be misalignment sometimes regardless of your best efforts. So what do we do if we have misregistration? Well, if you look at the overlay of PET on CT here for stress on the top and rest on the bottom, you can see that the anterior wall of the heart is out into lung space on both of these. More so on stress than rest. And if I look at the short axis slices here and the vertical long axis slices, I can see a little bit of a defect or a fairly big defect actually in the anterior wall of stress. Not quite as bad on rest. This is a worst case scenario. This is something that might lead me to call ischemia. When after correction, where I have used software to align these scans, and every scanner that I work with today has software to align this, some as much easier to use than others, you can see that this is basically a very normal scan with normal looking images for stress and for rest. So advanced scanning techniques, time of flight, digital PET, and how do we deal with motion? Time of flight. So the technique here is we detect a coincidence event. We measure the time difference based on the speed of light between when the two photons were detected. And we use this to calculate the location. Now, we can't get it down to a second. Now, we can't get it down to a centimeter, but we can get it down to several centimeters. And now when I'm reconstructing, I can limit the source of my photons to a smaller range instead of the full area inside the detector. That requires a fast detector. That requires circuitry with very accurate timing. Now, the next technique is digital PET, which provides a fast detector and improved time of flight. That maybe should be a faster detector than the one in the standard system. And here you can see the red line showing that I've narrowed things down a little bit. And over here, I've narrowed down the area where those photons might have originated even more. That gives me an improvement in my resolution. It means I can get by with fewer counts in my image. So how do I make a digital PET scanner? Well, all we mean by digital PET is that we don't have photomultiplier tubes or PMTs. We replace them with an array of solid-state detectors called silicon photomultipliers, or SIPM for short. SIPM for short. And here you can see how that array would replace the photomultiplier tubes. And so there's a space saving, but more important than that, the silicon photomultipliers enable me to work with higher count rates. They give me higher timing resolution. That gives me better time of flight. And in the end, I can do a more accurate reconstruction by using the digital PET. So this slide provided by James Case shows me differences between a dedicated PET scanner, a PET scanner that's similar in design, but now with time of flight imaging. And you notice there is a significant increase in signal-to-noise ratio, as well as an increase in image quality. And lastly, going to this gap where the largest sphere has been removed, a very nice looking image with a much, much higher signal-to-noise ratio. So this is because this scanner has digital detectors, improved time of flight, and also is a larger field-of-view scanner that gives me higher sensitivity. And that also contributes to that improvement. Motion management. So in the heart, we have all types of motion. We have the motion of breathing. We have the motion of the heart as it's beating. Deal with that with gated PET. But gross motion, breathing, patient may jump when they receive a pharmacologic stress. And then they may just shift their body as they become uncomfortable lying on the table for such a long time. This paper by N. Armstrong and others that is currently online at the Journal of Nuclear Cardiology to be printed sometime in the next few months deals with patient motion, and that's the gross motion of the body. And one thing they found is that in rubidium myocardial perfusion PET, 58% of the stress images and 33% of the rest images exhibited some motion. And the diagram over here, which is a figure from that paper, you have two patients that were rated as having some motion, two patients that were rated as having significant motion. And if you look at the short axis with no motion correction, with the motion correction technique that they created, you'll notice that you can get improved visualization of the myocardium, particularly true in the vertical long axis, but also better separation of the myocardium and extracardiac structures that you see right here. So motion compensations rapidly advancing through data-driven methods, better visualization of the heart, it gives you better separation of the heart and extracardiac activity. This slide provided by James K. shows a standard PET scan. This is a digital PET scan here with time of flight. But there I just changed to a technique called respiration compensation with motion freeze. What this tries to do is only give you PET at one point of the respiration cycle. So I'm not doing a full respiratory gated scan. What I'm doing is creating an image as if the heart was not moving at all with respiration. And you can notice between here and there, how much better the heart wall can be visualized and how it appears that you have better resolution because you've removed that component of the motion. Moving on to quality control, suggested procedures, and then the daily QC. Well, I want to direct you to the ASNIC imaging guidelines, not just for PET, but for SPECT and anything else that you do in nuclear cardiology. And for PET, their recommendations are a daily scan as recommended by a manufacturer, which might be a blank scan or a phantom scan, weekly or monthly sensitivity, and overall system performance. Accuracy, at least annually, scatter fraction, accuracy of attenuation, correction, image quality, and then other measurements. Every manufacturer has their own recommendations for what you should do, and you should follow those guidelines from the manufacturers stringently. We do these tests in my lab at least twice a year. So one way of doing that daily PET QC, which tests the performance of the PET scanner, similar to the scintillation flood of it, scintillation camera daily uniformity flood. I can use a uniform phantom, place that in the scanner, make an image of it daily, first thing in the morning. Then, and usually use germanium 68, which decays to gallium 68, which is a positron emitter. Or rotating rod source, where I have my rotating rod here, it's the same geometry as the rotating rod for attenuation correction discussed earlier. Some scanners have that, that's called a blank scan, because it's done with no attenuation, not even the table. Here you can see a daily flood, and the analysis shown for that. Here you can see the cyanogram of that uniform phantom. This is filled with epoxy that's impregnated with the radionuclide. Here you can see a map of the PET modules in this system. Down here, quantitative analysis is given. And in this case, everything passed. If something had failed, it would be highlighted in red, and the column here would say failed. For rotating rod scan, display of operational parameters is given. On several ways in the blocks that you see here. In this example, there are two defective detector blocks, and I know it's very small, but there's some areas highlighted in red where these blocks failed at the bottom. And this is a more overall system failure. And again, things that failed highlighted in red in the report. Very important this be done on a daily basis. And of course, if you have a PET-CT, then you have to do that CT quality control according to manufacturer's specifications. This says image quality, uniformity, noise, the thickness of the slice, and then the alignment of the system. So in summary, today we've talked about PET scanning, and that it requires coincidence imaging, dense fast scintillation crystals with high light output, ring detectors that surround the patient. Accurate attenuation correction is a must. The transmission scanning can be done with the rotating line source of CT. We have to minimize the motion for patient breathing, both in the PET and in the CT scan. We have to be careful to inspect the alignment, the alignment, the transmission scan and the PET as part of our analysis of each image. And we need to be able to realign if necessary. So PET scanning continues to evolve. We have improved detectors and techniques. And we need to adopt those technologies as we are able to do so. Lastly, quality control is important to our practice. We have daily procedures for both CT and PET, and we have to follow those procedures. We also have to make sure that we routinely do the periodic testing and calibration that's recommended and suggested by the manufacturers. Well, I'd like to thank you for your attention today. And with that, I'll conclude my talk.

instrumentation

cardiovascular PET imaging

PET scanner

coincidence imaging

attenuation and scatter

PET-CT scanners

quality control