Hello, my name is James Case, and I'll be giving a presentation on the physics of PET. Here's my disclosure slide. I'm an owner of CVIT that markets Imogen Pro and Imogen Q quantitation, and I received grant support from GE, Bracco, and DraxMash. The physics of PET has expanded tremendously in terms of our understanding of how to do this well and the technology available for us. I'm putting this one slide up as a resource that you can use to dig deeper into this. But really the key difference that we're going to have between SPECT and PET really comes from the fact that SPECT is a single photon process, and what that means is that a single photon is emitted and received and counted by the system. We can't bend gamma rays like we can regular visible light using a lens or a mirror. They have to be focused using a collimator. In SPECT, we use a lower energy photon between 70 and 165 keV, but really the key difference between SPECT and PET is really that attenuation correction is traditionally not used in most SPECT studies that are done. Now if we look at PET on the other hand, the way we create an image in PET is to identify the simultaneous acquisition of two 511 keV photons, so we'll say one here and one here, and by knowing where those are between a line of sight, we don't need a lens or a collimator to identify where those photons come from, and we can create an image without the tremendous account loss that we would have with SPECT. The other key advantage that we have is attenuation correction is very simple to do in PET by comparison to SPECT, but it's also very necessary to do so. All the images that we see in cardiac PET are going to be inherently attenuation corrected. Now at the highest level, virtually any good PET system or PET-CT system out there is capable of producing very high-quality images. The top row on this slide is from a dedicated 2D system. Underneath that is from an older but still a dedicated PET system in 3D, a roughly 15-year-old 16-slice PET-CT, and finally at the bottom, a more modern 128-system processed with time of flight. Each one of these are very high-quality images. So long as we understand the instrumentation that we're using and we follow the appropriate protocols and perform the appropriate QC, we're capable of doing high-quality cardiac PET work on virtually any of the machines available today. But the key point is to make sure that we understand the physics involved with creating a high-quality image. So the way that positron imaging works is to take advantage of the fact that when a positron annihilates in a material, it produces, it has to conserve the energy of the positron, which is a positively charged electron, and the electron that it comes in contact with and produce two photons traveling with the same center of momentum where they annihilated traveling in the opposite 180-degree directions. So what happens is the positron is emitted from rubidium or fluorine, and it goes through the medium and slows down by ionizing the water in the medium until it comes to a stop, then picks off a electron and then annihilates. The higher energy of the emission of that positron, the more it wanders away from its original site. But once it annihilates these two photons traveling in opposite directions, the way we take the acquisition is there's a timing window that within a very narrow time window, we try and capture the arrival of these two photons, then say, aha, that's a coincidence and that's what we use. Now, in addition to the trues, which is this set of data that's over here, there are three other type of events that a camera can see. The first is a single event. The single event is where one of the 511 KeV photons can be absorbed by the system, but we don't find its companion, so that's a single. And the higher the singles rate, that can have an effect on the dead time of the system. The second sort of event is scattering, a lot like scatter and spect. We have a, when the photon is emitted and it scatters off of one of the electrons within the medium, it will create a fictitious line of sight, lowering the contrast and resolution of the final image. A new type of event that we have in PET that we don't see in spect is something called a random. And a random is when we have two singles event that happen to occur just by chance within this narrow timing window of the scanner, and it creates, again, a fictitious line of sight that lowers the contrast of the images. Now, these systems, unlike spect, where we have one sheet, an angular system with one sheet of scintillator and behind it, a system of photomultiplier tubes, these systems are constructed of a number of small modules arranged in a ring around the system. So each one of these modules acts like its own little spect camera. What governs the properties of these is the detected material that's used. And what's very important is to have a high density piece of material, so that allows, the high density allows it to stop the photon inside of the material, a very high light output. Once the photon is stopped, it produces a lot of light, and that allows us to improve the energy resolution. And then a very rapid light decay, and that allows us to have a very short dead time. So as you can see here, one of the most commonly used one in PET is LSO. We're also very into that, LYSO and so forth, where it has a very short mean free path. So that means the photons don't travel very far into the material before they're absorbed. It gives up a lot of light, so 75 photons per 5.11 event, but it comes back to zero very quickly. And that allows us to have a very rapid reset up of the system to be able to acquire the next count coming in. Now, with attenuation correction, there are two different strategies that a scanner can do for attenuation correction. Over here on the left is a PET-CT system where we have the PET scanner, which is the thing that's doing our nuclear imaging, paired up with a standard CT scanner. And so inside of this crate would be the CT scanner. The other approach is to use a scanning radionuclide. So inside of this scanner, we would have a system of rotating radioactive rods, which then are used to create an anatomical image of the patient. Both of these systems are capable of producing a good attenuation map for doing the attenuation correction. For the line source-based attenuation, it's very simple. We see a projection of the activity through the patient, and what we end up producing is a set of lines of response. By one of the nice things about this, as opposed to SPECT, in PET, all we need is the actual image of the transmission data, and pixel by pixel, we multiply it by the attenuated image. We create an attenuation-corrected dataset for our image, and that's the nice thing. That's why I said earlier, it's a very simple process. We only need to multiply the transmission data times the emission data to create attenuation corrected projection data that we can then do our iterative reconstruction on. With PET-CT, it's a little more complicated. We acquire a CT volume, and the first step we have to do is dumb it down into the lower resolution as if it were acquired with the line source, and then forward project it to create the same projection dataset we saw on the previous slide, multiply it with our emission data. It's still very simple mathematically to do these things, so very highly reliable result for the attenuation correction, but a few extra steps for doing a CT-based. What we have here is three different images from different vintages of scanners. The far left is from a 2007 manufacturer year PET-CT scanner we can see. There's a fair bit of noise in the image, and the smallest rod source is still fairly ... It's a little blurry, but it's a lot of noise. As we move forward to a more modern system, we have roughly the same resolution, but a much improved signal-to-noise, as we can see in the table down below, and finally where we're at today is digital PET-CT, which we see a substantial improvement in both the signal-to-noise, almost a three-fold improvement, or over three-fold improvement in the signal-to-noise from the same phantom, and a much improved resolution. The smallest rod here is actually the smallest rod that we see in the other, so very clear images that we can get today with a digital PET-CT. Some of the advanced techniques that we have in terms of reconstructing these images is 3D reconstruction in time of flight. I want to briefly talk about these. 2D, when we think about cardiac PET as opposed to SPECT, we normally don't think about septa, but in a lot of the older systems in use today, they do have septa in them, and they're not as a part of the image formation process, but they're really used to almost act like a clutch on a car, to slow down the number of counts, reduce the dead time, and to limit the scatter. So on the image on the left, what we have is we can accept, because the septa creating essentially we're going to only accept counts within slabs of the image, and this allows us to exclude scatter, a lot of scatter, and a lot of randoms, so reducing the dead time, but we give up a fair bit of sensitivity, so we can reduce by almost tenfold the amount of scatter in the image, but it comes at the expense of almost a fivefold loss in the sensitivity of the system, whereas most modern systems today use a 3D acquisition. And here's the result. On top is a patient that's imaged in 2D with a 53 millicary dose, and the same patient acquired using a 3D study with half of the dose, easily, you know, as good if not better image quality using the 3D versus the 2D. The second major improvement that we have in reconstruction is the use of time-of-flight reconstruction. In a traditional reconstruction, we don't have any idea where a photon was emitted along a line of sight, so we make an assumption that it has equal probability of being emitted anywhere along this line of sight. But using the timing of when one photon might arrive at one side of the scanner versus the other side of the scanner, we can narrow down the probability of where that photon could have originated from, and that's time-of-flight reconstruction. It allows us to have higher resolution for the image, better uniformity of resolution across the field of view, and probably what's most interesting for what we do in Cardiac Pet, better object differentiation. So if we have a hot bowel on one side and the heart on the other, we see less crosstalk between the two in a time-of-flight reconstruction. So here's an example of the power of time-of-flight. So this is a digital image of a patient off a digital PET-CT system, but using a traditional iterative reconstruction. By contrast, if we use the time-of-flight reconstruction, you can see much better definition of the walls. We can also improve these using diaphragm and breathing corrections and other motion corrections. So this is the optimal reconstruction that we can achieve when we add all of these corrections onto our reconstructions. Now moving on to the tracers that are available, we really have a fairly limited number of tracers at our disposal for doing Cardiac Pet. Rubidium-82 is probably the most commonly used, it's a potassium analog, very similar to thallium, has a very fast half-life, and it's delivered by a generator. Now the second most common is N13 ammonia, very good flow characteristics, but because of its 10-minute half-life, it typically has to be produced on-site. And finally, in Phase III development right now in the United States, there's a new F18-labeled perfusion agent called flirpyridaz, and there will be a lot of discussion about that at this meeting. So how does the strontium-82 work? The way it works is that a saline is drawn across a strontium generator, and the strontium generator is constantly producing new rubidium-82, and the concentration builds up inside of the generator. When we push the button to start the generator, it flows the saline over the column, it grabs that rubidium that's been produced, passes it through a system that does a radioactive assay, and then delivers it on to the patient. Now the production of N13 ammonia is a little different. The way that it's done is it's done with a cyclotron, and what a cyclotron is, is it is a system of electromagnets that are used to accelerate protons to a very high speed so they crash into an atom and change the atom's nuclear content. So with these, what we're going to do is we're going to produce a stream of these protons to modify the ammonia so that it becomes its radioactive counterpart. Now to create ammonia, there are really two processes we use. Probably the purest way of doing this is using methane gas. But more commonly, what's been used of late is a cocktail of ethanol and water and putting that in the target range to produce the N13 ammonia. And as I said earlier, the new kid on the block is for Pyridaz. It's a product from Lantheus, and this binds is actually a derivative of Rodenone. And the way this works is an F18 is tagged on this mitochondrial one complex to give it the flow characteristics that we want, very nearly approaching the same flow characteristics of O15 water. So very, very attractive agent, still not available commercially, but in phase three trials right now in the United States. The radiation dosages that we see within PET, one of the nice things about PET is most of the agents, because the short half-life of F18 N13 ammonia and rubidium 82, we typically see much lower radiation doses to the patient without sacrificing the image quality. So here's a table of common procedures that are done, both SPECT and PET, but probably an easier way of looking at it is in this fashion. And what we have here is SPECT protocols are laid out in red, and PET protocols are laid out in blue. And as you can see, the lower dose protocols are all our PET protocols, and then the SPECT protocols, with the exception of the CCT-based stress-only protocol, are going to be of a higher dosage. So when we apply these PET protocols, not only are we achieving higher image quality, better flow characteristics, we're actually simultaneously reducing the exposure to the patient. So here, I'm not going to go over this slide. This slide is some of the recommendations from the American Society of Nuclear Cardiology for minimizing dose. The reference is at the bottom, and this will be contained with slide sets. So artifacts in PET, they all come from kind of a similar fashion that we might see with SPECT, patient motion and camera setup, but they manifest themselves in a very different way. Early on, I pointed out that all of the studies that we're going to be looking at in PET have attenuation correction. And because of the fact we do attenuation correction, we have an anatomical image that we go ahead and use for doing the attenuation correction. If that anatomical image doesn't line up properly with the emission data, we can get artifacts. This will be discussed later on. We also have motion caused by CT. The patients are asked to either hold their breath or free-breathe, and if the diaphragm is caught in multiple locations, it can introduce artifacts in the images as well. And not only can we have artifacts introduced because of the quality of the transmission map, which is used for attenuation correction, we also can have motion within the emission data, as you can see in the bottom, or intra-scan motion as opposed to inter-scan motion. So here's probably the most common artifact that we would see within cardiac PET. We look at this type of a study, and it would be very hard to call this a normal study. This is one of the classic examples of misregistration. You can see that a fair portion of the heart is overlapping the lung field because the lung has a lower density than the soft tissue. We get less attenuation correction than we would expect, and we undercorrect the counts in the heart and create an artifact. And when it's correctly positioned, you can see this is the same study that we had over here. It's clearly just a very straightforward normal study. So it's very important that we maintain proper positioning of the transmission and emission data sets. So how important is this? Roughly 21% of resting studies have misregistration. As little as one centimeter is necessary to introduce an artifact. And in PET-CT studies, as high as 40% of studies can have a false positive result as a result of misregistration. And because of that, whenever you're doing cardiac PET and PET-CT, all studies have to be inspected for misregistration and corrected when possible. I mentioned inter-scan motion. That's motion within the emission study itself. We usually identify these sorts of artifacts by the matching 180-degree artifact. You can see on the images below, they're 180-degree artifacts. And what happens is, is that the image is smeared in the direction of travel. And so perpendicular to the direction of travel, there's less blur. And so we maintain the counts, and in the direction of travel, there are fewer counts because the tissue isn't overlapping. And you can see we end up with these 180-degree artifacts. What causes these is the patient falling asleep. They can breathe heavily during the stress because of discomfort. They can cough. Talking will do this. So again, with these patients, when you're imaging them, make sure that they stay awake, they keep quiet, and you make them as comfortable as possible. But again, very important they don't fall asleep and to keep them still. For PET-CT, breathing protocols are very important. There are several different approaches for doing it. But both of these studies are normal studies. You can see very interesting artifacts can be introduced if the diaphragm is imaged in multiple places. The way you identify these is you can see in the example above, something that looks a bit like a mushroom sitting over the liver, where the liver is seen in two different places, or this little notch artifact where there's almost like a cough right in the middle, and it can introduce a little notch in there. So finally, in summary, most systems, so long as you follow the protocol and understand the science of the system that you're using, are capable of producing good quality images. The acquisition protocols, make sure you have a structured protocol that defines when you're going to be imaging so that you don't capture too much of the blood pool. Make sure you have good counts between the transmission and emission. Before you start interpreting the study, you always make sure that you have proper acquisition. You've inspected for motion, the quality of the infusion, and the processing. When you see misregistration, always send that back for reprocessing and correction of the misregistration. And finally, in the interpretation, be sure to inspect the data and make sure that everything is correct before you dive into interpretation. Thank you.

James Case

PET

positron emission tomography

nuclear imaging

511 keV photons

attenuation correction