Hello, my name is Rob de Kimp. I'm an imaging physicist at the Ottawa Heart Institute in Canada. The topic of this lecture is quantifying myocardial blood flow using PET, an overview of the nuts and bolts. These are my disclosures, mainly related to rubidium PET imaging technology, generators and software. So why are we interested in myocardial blood flow? Well, here you see an example of results of a relative perfusion scan that we typically view, for example, as a left ventricle polar map with some template coronary arteries overlaid. When there are no obstructions in these arteries, we see a relatively homogeneous uptake of the tracer with the representative of normal and no obstructive disease. This is effective when we have potentially an occlusion or a narrowing in one vessel, we can see reduction in tracer uptake relative to those normal areas. But in cases with multi-vessel disease, three-vessel disease, or even microvascular disease in the small vessels, the relative appearance may look homogeneous, although this heart is typically not have homogeneous or adequate blood supply. And this is one of the reasons to investigate or be interested to use quantification of myocardial blood flow, where we can now attach quantitative units in terms of mils per minute per gram on a more quantitative physiological scale. We can see now that the flow in this left ventricle is homogeneously reduced and clearly abnormal. The imaging protocol that would be used for blood flow quantification would be very similar for most tracers compared to what's done for SPECT for fusion imaging. You would start in the case of CT with a positioning scan, a scout or a topogram to localize the heart, followed by an x-ray CT scan that's used for attenuation correction. With PET, this is done in every single case. It's an integral part of the imaging procedure. We would then at baseline inject our PET tracer and scan for 5 to 15 minutes, depending on the tracer that's injected. And currently, there are three approved perfusion tracers in the U.S., rubidium, N13 ammonia, and recently, F18 for pyridines. This is followed for those studies where blood flow quantification is desired, can only be done currently with pharmacologic stress testing or coronary vasodilation with one of these vasodilators, regudenosone being the most commonly used again in the U.S. And at peak hyperemic stress, we would re-inject the tracer, scan again for the same length of time, and sear the goal of unmasking regions with potential perfusion defects. And depending on the tracer and the pharmacologic stress agent, this can be all completed in 20 to 30 minutes. The basis of blood flow quantification is based on the recording of dynamic or cine-dynamic PET scans. And here, you can see an illustration or a video showing you the time course of uptake in the thorax, starting from the time of injection. You see the tracer in red in arterial blood, corresponding to the location of the crosshairs. And over time, the activity is cleared from the blood and taken up into the artesian. And it's, in fact, this transition from arterial blood as a shape of the valcardial curve that's the basis for quantification of blood flow. We created a simple benchtop model to illustrate the properties or the nuts and bolts of how blood flow quantification really works. So here, if we had a pump that was pumping, let's say, saline in through a capture column to a waste bucket, and we had some activity that we were interested in quantifying the rate of flow, and some magic capture column that could convert our red activity into blue activity after being captured. If we flip this valve and allow our activity to be pumped into the capture column, we would see that over time, we have a flat or a square wave injection. The activity in our capture column would increase linearly over the time of this tracer injection. We could then continue a measurement period from, let's say, one to six minutes to capture the measurement of that activity. And in this case, if we have our input activity, let's say, arbitrarily at a thousand units, and we have performed that dilution for one minute at a flow rate of 0.5, the amount of activity that would be captured in this column would be a thousand times 0.5 or 500 units or better else in this case. We can see if we rearrange this equation, I'll flip back to the basic explanation. The captured amount is product of blood input area and flow rate. If we rearrange this equation, we can invert it and say flow rate can be calculated if we know the amount that was captured after some stable input divided by the area under this blood input current. Illustrating how this would change if we increase the rate of our pump, we will see that for the same blood input, we now are pumping more tracer or more contrast through our capture column and will result in a higher stable output value. And again, using our equation, we can calculate what this flow rate would be. We now have four times as much activity captured. We've increased the flow rate from 0.5 to 2. We have the same input area still at a thousand, and we can deduce that the flow rate in this case was two mils per minute. This works for a number of different flow rates, and it also works for a number of different tracer injection speeds. And so here you can see on the top, we would maintain the area under the blood input curve at one minute, but we would change the duration. So in this case, it would be tracer would be ejected relatively quickly over a half a minute, over one minute, the width is wider over two minutes. We're spreading that over a larger period, but in every case, because the blood flow is the same on each row, the amount captured by this column remains the same. As we increase the blood flow from 0.5 to 1, the amount captured increases and increases further as we increase it to 2. So a common question with respect to blood flow quantification is, does the ejection profile really matter? It's not of critical importance. Variable blood shapes, in fact, are okay. Now, if we move from our benchtop model to something a little more physiologically relevant, we would replace this pump in the capture column with something like the heart, which is actually pumping blood through the systemic circulation, but also through the coronary arteries. And in fact, it's the pre-arterials that are regulating the flow rate being delivered to the myocytes in response to demand. If we now activate this valve, inject our tracer, we can see that the tracer goes in, you know, maybe something a little more like the variable peak, but the amount retained in the heart still depends on the area that was supplied under that blood input curve. And the relationship is exactly the same. In this case, we've simply added now concentration of tracer per gram, and we can measure cardiac effusion in mils per minute per gram of myocardial tissue. What I've explained is actually the exact basis for what's called a retention model for quantifying blood flow. In this case, you see an example of blood time activity curves that was measured in the LV cavity, and in blue, the tissue response. And with this simplified retention model, often depicted here, where we simply have assumed that the tracer moves from material blood into tissue and is irreversibly tracked. And in this case, what's required is simply to make one measurement of the tissue activity at the end of the scan. We measure the area under the blood curves, take that ratio, and gives us an index of flow. There are some tracer and scanner-dependent corrections that are required, which you can read about retention fraction and partial volume corrections in this paper cited here. A more physiological model is called compartment model. This is a physiological model of tracer exchange, where we realize that tracer may move from artery into tissue, but then also leaves the tissue and be carried out in the penis outflow as well. And in this case, we have a little bit more flexibility in terms of the shape of the blood input and really characterizing the transformation of that input shape in red to the full shape of the output curves over the full time course. And in this case, we need something a little bit more complicated than simple arithmetic division. We use a mathematical procedure called deconvolution, but the concept is the same in that the incense ratio or the deconvoluted curves give us an index of flow, again, with some corrections depending on the tracer properties, initial extraction fraction in this case, and partial volume correction depending on the scanner resolution. The basic steps towards blood flow quantification are illustrated here. Then we have our imaging. The process starts with segmentation or identification of the left ventricle shown in blue. This is our tissue response for the output of the system. We would position regions of interest either left atrium typically or the left ventricle to get our input function. And in this case, we show a compartment model fit illustrated by the solid line that is fitting to our measured data points that were acquired during our dynamic scan. Then for quality, really, it's very important to capture the entire shape of the blood input of a tissue response starting from zero. And of course, with any model fit, you should verify that your model, i.e. the line, is fitting your data points to have confidence in the measured flow values. And as a result of this, we typically get polar maps, again, because we have only pulled out the left ventricle from these measurements. And then our primary measured parameter is blood flow shown in the second polar map, top left. There are, you can see that contrast is improved compared to uptake only, and this image is more representative of the retained tracer. This is representative of the blood perfusion that delivered that tracer to the myocardium. These outputs will vary depending on the image quantification software that you're using, and there may also be some quality assurance metrics to help determine whether the quality is sufficient for clinical interpretation. Having done this, we're under resting conditions and under stress conditions. Typically, there will be also a ratio calculated in terms of blood flow reserve, and this is the stress-rest ratio. In some cases, it may also be useful to look at the difference, which can be a little bit more sensitive way to look at coronary steel, where flow may actually decrease at stress compared to rest. There are some normal limits for, let's say, gray zones. Normal flow, but it's common to have abnormal flow, the flow reserve, below some gray zone, below 2, stress flows below 1.5, delta values eventually below 0.75. Some have also introduced the concept of coronary flow capacity, where we look at, on the bottom right scan, stress flow together with reserve, indicating that both stress flow and reserve would need to be abnormal to have abnormal flow capacity, shown here as the cyan or the light blue. There are several quality assurance steps that are very important to understand and to learn about when starting blood flow quantification. As I mentioned, it's based on first-pass imaging, so it's very important to start the scan at the time that the tracer arrives. FIT-CT alignment for attenuation correction is as important for flow as it is for conventional perfusion imaging. Perhaps the most important component to be aware of in the quantification of blood flow is to evaluate whether there is patient body or heart motion. Most software programs will now include patient body motion correction in order to get accurate blood flow estimates. Also, a number of factors can affect the particular values of flow that have been validated in the literature. These are a little bit beyond the depth that we can get into in this short presentation, but if you are starting flow, I'd encourage you basically to look for publications and follow, where possible, a validated acquisition protocol and analysis pipeline. And with that, I'd like to thank you for your attention and thank the ACC for their support of this educational program.

myocardial blood flow

PET imaging

tracer retention

compartment modeling

quality assurance