Hello, I'm Paul Kremer, a cardiovascular imager at the Cleveland Clinic, and it's great to be with you for this curriculum on PET protocols. So today I'm going to discuss how to perform a rubidium-82 PET myocardial perfusion study. I have no disclosures. So we have several learning objectives for this lecture. The first is to describe how to prepare patients. Next we'll learn procedures to optimize rubidium-82 injections, and finally we'll learn how to acquire and process rubidium-82 cardiac PET images. So first, how to perform a rubidium-82 PET myocardial perfusion study. So when we think about camera setup, most dedicated PET cameras consist of rings of small detectors, a few millimeters on a side, tens of millimeters in depth. And coincidences between detectors in a single ring produce one tomographic slice of data. So as a refresher, we know that there's different types of detections. There's the true detections, where the 511 keV photons after an annihilation reaction hit the detector at the same time. There's random detections, which importantly can be counted as coincident events. Then there's detections related to Compton scatter. And finally, spurious detections. Now there are 2D septa-in PET scanners, where you have the septum, which is usually lead or tungsten, between rings. And the septum shields coincidences between one ring and distance ring. And this dramatically reduces the scatter events that are detected. Most modern scanners are going to be 3D or septa-out PET scanners. This allows coincidences between all possible rings, so it dramatically increases the sensitivity and allows us to give a lower dose. But importantly, it does increase the scatter and increases the count rate with a decrease in dead time and random events. So the next thing about camera setup is thinking about crystals. And we have three different crystal types, common for PET, ESO, excuse me, BGO, GSO, and our lutetium-based crystals, LSO or LYSO. And the main advantage of GSO or lutetium-based crystals is reduced dead time. So when we think of our camera setup, in particular as it relates to quantitative or bitum82 scans, we have a wide range of count rates. So 3D acquisitions can be problematic, as I mentioned previously. However, the GSO and the lutetium-based crystals have advantages during these high count rates after rubidium-82 infusions. And in particular, GSO and lutetium-based crystals can acquire data and minimize effects of randoms at higher count rates associated with 3D imaging. And as I touched upon earlier, this allows us, in general, to give lower doses. And you can see based upon whether you're using 2D or 3D and an LSO, GSO versus a BGO crystal, your dosing may range anywhere between 10 to 60 millicuries. And in general, the doses are going to be lower on 3D systems with lutetium or GSO-based crystals. So what about PET quality control? So we perform acceptance testing upon delivery and with any major hardware upgrades. And as far as our daily QC scans, these are as recommended by the vendor, but generally consist of constancy and uniformity. And every day before we start scanning our lab, we perform a blank scan. And sensitivity testing is generally performed every week. In terms of accuracy, so testing for corrections for randoms and count losses, this is performed at least annually, as is the assessment for scatter fraction. And in terms of the accuracy of attenuation correction, this is performed at least annually, as is an assessment for image quality. Okay, so what about attenuation correction? So of course, cardiac PET is only performed with attenuation corrections. And there's two techniques for transmission scanning. You can have line source, for example, with cesium, or patient-specific attenuation maps. And these transmission scans are acquired sequentially, either pre or post injection of the radiopharmaceutical, and it's essential that the patient remains still during this time. So that's for our PET quality control. What about the CT aspect, as it relates to our study? So for CT quality control, we look at calibration, which is the accurate absolute CT numbers in Hounsfield units, as quantitative CT values are used for attenuation coefficients at 511 KeV. And this is generally performed monthly. And CT calibration is checked daily with a water-filled cylinder. And then we also assess for field uniformity in our CT quality control. So the reconstructed image or a uniform water-filled cylinder must demonstrate low variation in CT number, and this is typically performed monthly. So for the PET CT quality control, we look at registration. So that is the reconstructed PET and CT images must reflect the same 3D location. And any errors after software correction should be less than one millimeter. So this is something, what I look at, the first thing in every scan that I'm looking at is the registration. And the attenuation correction accuracy. So it's important that an image with a water-filled cylinder to assess the PET field uniformity and PET activity concentrations after CT-based PET attenuation correction. Okay. So that's some of the quality control issues. Let's get into the specifics of a rubidium-82 protocol. So the patient preparation is the same as for SPECT. So that is, for example, fasting for at least six hours, no caffeine for at least 12 hours. The patient is generally supine with arms above the head. So the rubidium-82 is eluted from a strontium generator in 10 to 50 mLs of normal saline. And this generator is fully replenished every 10 minutes. And at the beginning, a scout scan is typically performed before every injection to localize the heart. And then a CT for attenuation correction is performed. So it's notable that the current on this CT is typically low, so 10 to 20 milliamps. And in general, we're not adjusting the KVP the way we may do with other scans such as cardiac imaging, for example, coronary CT angiography, so typically at 120 KVP. Now, we typically perform LIS mode acquisition. And these data are then binned into the gated static and dynamic datasets for subsequent analysis. And images are reconstructed with filtered back projection or with iterative reconstruction. Now, after the injection, about 80% of counts are acquired in the first three minutes, 95% in the first five minutes, and 97% in the first six minutes of the study. With a normal left ventricular systolic function, static imaging starts 70 to 90 seconds after the injection. However, you may want to delay that static imaging acquisition in the setting of left ventricular systolic dysfunction. So for example, in a patient with mild to moderately reduced left ventricular injection fraction, static imaging can start 90 to 110 seconds after injection. And with severely reduced systolic function, you can push that out even longer to 110 or 130 seconds. And the reason we do this is to reduce excessive blood pool counts. So I may be reading a study in a patient that has a very low LVEF, and there's a lot of signal in the blood pool, so I'll go and ask the technologist, hey, can you push this static acquisition out another 30 or 60 seconds and reprocess that for me? Okay, so I'm going to touch briefly upon the CT protocols just to contrast them for what we do for CT for attenuation correction and a standard Rabidium 82 protocol. So we do have protocols for hybrid imaging, but it's important to remember that the CTs we're doing for attenuation correction are non-contrast, non-ECG gated, free breathing scans, so about five millimeters of slice thickness. This is in contrast to a CT for coronary artery calcium scores, which is of course ECG gated, which involves a breath hold and is reconstructed at three millimeter slice thickness in comparison to coronary CT angiography, where you give intravenous contrast, it's ECG gated, it's breath hold, and typically with sub-millimeter slice reconstruction for thickness. So as I touched upon, the CT for attenuation correction has a lower tube of current for a transmission scan. We typically acquire this at end expiration breath hold or with shallow free breathing. We don't want to have a deep inspiration breath hold because that will not be what was the state of the patient during the admission scan, and it's important to recognize that metal artifacts can cause problems with our reconstruction algorithms such as ICD leads. So CT for attenuation correction, the slice collimation and reconstruction thickness should be the same as our PET, so generally around four to five millimeters of slice thickness. Again, this is in contrast to coronary artery calcium scan, which will be three millimeters of slice thickness, or CT coronary angiogram, which will be 0.5 or 0.75 millimeters. And ECG gating is also not recommended for attenuation correction. So coronary artery calcium scans are typically prospectively ECG triggered, acquired in mid-diastole, and generally, as I mentioned, during an inspiratory breath hold. So this may not register with your MPI for attenuation correction, so something to keep in mind as well. And as it relates to coronary CT angiography, if it's combined with myocardial perfusion imaging, it's preferable to perform the MPI first to minimize potential interference in attenuation correction from the iodinated contrast that's been injected. And with coronary CTA, you need a large bore IV if the patients are prepared with beta blockers and sublingual nitroglycerin. Now with coronary CT, we do adjust the two potential based on body habitus to optimize the image and minimize the radiation dose as much as possible. Retrospective triggering also reduces radiation. And if retrospective gating is used, perhaps because of an arrhythmia or faster heart rate, we typically modulate the radiation dose to further reduce the radiation exposure. In coronary CT, the scan length is typically from the carina to below the inferior portion of the heart. And the contrast timing, the bolus scan involves timing to peak capacification of a test bolus. Or more commonly, we use bolus tracking acquisition, which can be triggered on the scan based on pre-specified enhancement in the aorta, for example. Once the bolus reaches an attenuation, say, above 150 ounce field units, then that will trigger the scan. And then we can reconstruct to the thinnest slice as possible. Okay, so that's just to really highlight the differences between what we're typically employing with a CT for attenuation correction versus some dedicated CTs for coronary artery calcium scoring or CT angiography. Now to touch upon quantification of myocardial blood flow, which has become a standard assessment for a lot of the patients that are being scanned in our lab. And quantitative myocardial blood flow is the absolute hyperemic myocardial blood flow in mLs per minute per gram of myocardium. And myocardial flow reserve has been shown to improve diagnostic accuracy, reduce false negative studies in the setting of balance ischemia, and improve risk stratification for our patients. So where are the scenarios where in our lab, or generally speaking, quantification of myocardial blood flow is most useful? So one, to identify patients with an increased suspicion for multivessel disease. So if you have a perfusion defect that's in a single territory, but the myocardial blood flow assessment is quite low, then you'd be more suspicious for multivessel disease. Or if you even have relatively normal, relative perfusion and severely reduced flow, that can be a patient, as I mentioned, who has balanced ischemia, quote unquote balanced ischemia. And in certain patients, more and more to identify microvascular dysfunction. So when there's a disparity between the perfusion results and coronary angiography, CT-based or with invasive coronary angiography. And then finally, to assess cardiac allograft vasculopathy in patients' status post-heart transplant. So those are a few examples where I think quantification of myocardial blood flow is most helpful clinically. So conversely, what are some scenarios where it may not add diagnostic value? So patients who have had prior bypass surgery who have diffused reduction in flow despite a patent grasp, in patients with large transmural infarcts with severely reduced resting flow, in patients with severe LV dysfunction, also patients with severe valvular heart disease. The other category that I think should be considered where it's of unclear diagnostic value is in patients with severe renal disease. So those are some clinical scenarios where the quantification of myocardial blood flow may not add diagnostic value and may not, you know, in our lab we typically don't report the values in these clinical scenarios. So the quantification of myocardial blood flow, the images are acquired in dynamic mode with LISMOD acquisition, as I mentioned previously. The regions of interest are placed on the LV myocardium and blood pool and copied to acquired images to create myocardial tissue and time activity curves. Now these time activity curves are corrected for spillover from blood pool and radiotracer decay, and they're fitted with a validated tracer model. So Ritm82 kinetics can be described with one or two compartment model fitted using arterial input function and myocardial time activity curves. And I'll show a couple examples of what that looks like in clinical practice. So I think it's worth mentioning the technical factors that may affect myocardial blood flow quantification accuracy. The first can be related to our scanner and the spatial resolution of our scanner. So that can result in spillover or partial volume errors if there's limited spatial resolution. Another issue with scanner, which we touched upon, is data corrections related to dead times or randoms or scatters, which can result in quantitative errors. And then if you're using time-of-flight capability, that can result in imaging quality differences between different scanners. The next consideration is the technique. So for example, the infusion characteristics, whether you give a bolus versus a sustained infusion. The bolus depends on the age of the Ritm82 generator and the intravenous line quality. So for example, there's a possibility of double peaks on the time activity curves if the bolus is not a tight bolus when it's administered. Timing relative to scan start. So it's important that your technologist start the scan prior so that you don't miss the bolus and miss some of the counts, of course. The CT acquisition, which we touched upon in terms of breath hold or free breathing and attenuation-related quantitative errors and artifacts. And then the protocol in terms of a single CT for both rest or stress versus separate CTs can result in attenuation-related quantitative errors and artifacts. There's also issues related to image generation. So the reconstruction algorithm and parameters. So noise characteristics and convergence errors can result in potential bias in a myocardial blood flow assessment. Time framing. So the timing of the frames can also affect our myocardial blood flow. And then PET-CT registration. So the capability and quality of registration correction. So misregistration can result in attenuation-related quantitative errors and artifacts. And then finally, motion correction. So real-time or after processing can result in attenuation-related quantitative errors and artifacts. The next thing to think about is the actual analysis of the myocardial blood flow itself. So the software that's used that can be vendor dependent and there can be differences related to that. And in particular, how motion correction is performed. Whether your software allows you the ability to inspect and correct motion frame by frame in a manual fashion or whether it's done automatically. For example, motion can result in myocardial creep with artificially higher counts in the RCA territory. And I'll show an example of that at the end of the talk. The myocardial blood flow algorithm. So there can be differences between flow extraction models. So you need to know what your software is performing, of course. And with Rabidium-82, there can be errors at high flow rates in particular compared to N13 ammonia or O15 water. And then finally, how we define our region of interest in the blood pool. So these differences can result in quantitative errors. Okay. So I'll conclude with just some case examples. So first, this is just a normal rest stress Rabidium-82 PET. So you can see the stress on the top, rest on the bottom, and our short axis, horizontal long axis, and vertical long axis slices. And this is a normal rest stress Rabidium-82 PET myocardial blood flow assessment. So here you can see our time activity curves. This is the left ventricular blood pool. And these are various segments of the left ventricular myocardium. And we can see in this patient, the resting global blood flow is 0.9 mL per minute per gram, which augments to 2.0 mL per minute per gram with a flow reserve of 2.3. So here's an example of misregistration of the CT for attenuation correction in the Rabidium-82 emission scan. So here you can see that the CT of the lateral wall of the left ventricle is not well aligned with the Rabidium-82 emission scan on this axial and coronal slice. And that will lead to misregistration and CT attenuation related artifacts. And here's an example of something I mentioned before. This is excessive blood pool in a patient who has severe left ventricular systolic dysfunction. So you can see that here on this resting scan that there's a lot of background signal in the left ventricular blood pool. And what can be problematic is if you're doing a rest stress study and you give the patient a vasodilator and their heart rate increases, their blood pressure decreases a little bit, the cardiac output is augmented, and there's differential blood pool signal between the resting and stress images. So that would be a setting where I would ask the technologist on the resting scan to delay the static acquisition to try and match the blood pool signal between the rest and stress images. And then finally, this is an example of diaphragmatic creep with spuriously increased inferior wall myocardial blood flow analysis on the stress images. So what you can see on these stress and rest time activity curves is that here when the radio tracer is in the left ventricular blood pool, we see a lot of spillover into the inferior wall here in red, in particular on the stress images, which you can see and appreciate here on the polar maps. The result of this is a spuriously high stress flow in the RCA distribution. Now as I mentioned, there's different ways to do motion correction, whether the vendor provides it in an automated fashion or allows you to actually go through frame by frame and then reconstruct the contours of the left ventricular myocardium. And you can see after performing motion correction that the time activity curve of that inferior wall is now similar to the other myocardial segments and that our RCA stress myocardial blood flow, which was initially 3.3, is more accurately actually at 1.3. So it's very important anytime you're performing a PET and you're reporting quantitative myocardial blood flow, or actually any study just to look for motion artifacts, to really look at these time activity curves and if you have the opportunity to look at the data frame by frame. Okay, to summarize, GSO and lutetium-based crystals can acquire data and minimize effects of randoms at higher count rates associated with 3D imaging. Motion scans are acquired sequentially, either pre or post, and it's essential that patients remain still during this time. With registration, the reconstructed PET and CT images must reflect the same 3D location and errors after software correction should be less than one millimeter. So with our CT for attenuation correction, the slice collimation and reconstruction thickness should be the same as PET. So, for example, between four to five millimeters in contradistinction to CTs we may perform for dedicated cardiac imaging such as coronary artery calcium scoring or coronary CT angiography. In terms of quantitative myocardial blood flow, it's most useful in patients with an increased suspicion for multivessel disease, a concern for microvascular dysfunction, or cardiac allograft vasculopathy, and I highlighted some examples where I think we should avoid clinically reporting a quantitative myocardial blood flow. Notably, errors in quantitative blood flow can occur related to the scanner, related to technique, related to how the images were generated or analyzed, and so it's important to be aware of these factors if you're reporting these values in your daily clinical practice. Here I've provided some key references that I would encourage you to look at as well. So thank you, and again, it's been a pleasure to be a part of this curriculum.

rubidium-82 PET study

camera setup

myocardial perfusion

CT protocols

quantifying blood flow

diagnostic accuracy