Hello, everyone. My name is Ian Armstrong. I'm a consultant medical physicist working in Manchester, nuclear medicine department in the UK. And this talk will be an overview of the general physics that we would need for CT, particularly with a focus of what we're doing in nuclear cardiology for attenuation correction and coronary calcium scoring. In my disclosure, this is an overview of the talk that I'm going to give. So I'll touch briefly on how the x-rays are produced in the CT tube. And then I will talk about the acquisition and the reconstruction parameters that we can change as a user and the impact this will have on radiation dose and image quality. I'll touch on the calculations that we need to do and what parameters we need to look at when we're interested in patient dosimetry. And then finally, I will go over the two main applications which we're interested in for hybrid imaging in nuclear cardiology from a SPECT and a PET perspective. So let's start at the beginning. This is when we're looking at x-ray production, which we will see in a CT scanner. So a CT scanner has an x-ray tube. Here's an example picture up here. And this is a schematic diagram of what we have. So we have this cathode here and we pass a current through this. This is what we would call our MA. And this creates a cloud of electrons here. What we call sometimes is we boil the electrons off. This red block here, this is our anode. And we apply a positive voltage to that. And as the electrons are negatively charged, they are accelerated towards this anode. When they strike this anode, then they stop. And when a charged particle stops, it releases something called Bremsstrahlung radiation, which is a broad spectrum of radiation. And this is emitted in these green directions here. So what we see coming out of our x-ray tube is an energy spectrum that looks a little bit like this. So this is one of the main differentiating factors between an x-ray energy spectrum and an energy spectrum that we would expect to see in single photon or positron emission tomography, where those imaging modalities have a photo peak, which is according to the primary gamma ray energy, which has been used in the imaging. What we have on an x-ray is a monochromatic energy spectrum. We've got many, many different energies according to the level of Bremsstrahlung radiation that has been emitted. These will go up to a maximum energy. And you can see here, this spectrum is stopping at 100 keV. So this is what you would get as a maximum energy from an accelerating positive voltage of 100 keV on your x-ray tube. So for that reason, I typically, and most of the people who work in CT will refer to the x-ray tube voltage as the KVP, the peak killer voltage, because this determines the peak energy that you will see on your x-rays. These notches sticking out here, these are called characteristic x-rays, and these are a property of the interaction of the electrons with some of the atoms, the atomic electrons in the anode. And this is just another thing that you will see from an x-ray spectrum. This is what a CT scanner looks like with its covers off. You can have great fun Googling CT scanner rotating on YouTube, and you'll see these things spinning around at an enormous rate, typically getting up to about four times a second now. So what we've got here is we have our x-ray tube on the right-hand side, and we've got a detector here. And a CT detector, or a CT system rather, sorry, it will emit a fan beam of x-rays, as we can see by those blue lines here. And then this whole apparatus spins around. So we effectively have a rotating fan beam of electrons, of x-rays, sorry, as the system rotates around. The x-rays, as they pass through the patient, they are attenuated at different amounts according to the density of the matter that are passing through. So the detectors are effectively measuring the attenuation at a given trajectory as they go through on that fan beam. So like SPECT and PET imaging, we have our raw data and then we perform image reconstruction and we end up with our three-dimensional image volume with a series of voxels. Now the image voxels are an indicator of attenuation. That attenuation is normalized to water. And we refer to these voxels as Hounsfield units. So when we're normalizing to water, we can see here that our water has a value of zero. Anything less dense than water would have negative Hounsfield units. Anything more dense than water would have positive Hounsfield units. And these example numbers here indicate what we would see in various biological tissues and structures. So we'll move on now and we'll talk a little bit more about the parameters which we can change on our CT tube as a user and what they will do. So the first one we can do is adjust the tube voltage. And this is done in fairly discrete steps. Typically, an x-ray tube may be able to emit at 140 kVp, 120, 100, and maybe 80 as well. So in very much the same way as gamma rays, the higher the kVp, which is the tube voltage, the higher the maximum energy of our x-rays and therefore they are more penetrating. But because they're more penetrating, there's less differentiation between different densities in the tissue. And hence what we see actually is that the tissues are actually lower contrast. So we see smaller differences in the Hounsfield units as we increase our tube voltage. The tube current, that's a measurement of how many electrons are being emitted from the cathode per second and hence translates directly into how many x-rays are going to be emitted per second from the tube. And very much like emission tomography, when we're interested in the number of gamma rays that we collect, it's the same with CT. It's the number of x-rays that we can collect. This will determine image noise. And so the higher the tube current, the lower the image noise. We've seen already that that CT tube will rotate and sometimes you'll have a parameter, it could be called the exposure time or the rotation time. So this is how fast the tube is rotating and this can have various impacts on things. Mainly it will be how long the scan takes to acquire. So the typical time it will take to cover an organ for the entire scan range. So this, if you've got, for example, the heart beating, if you cover it very, very quickly, then you're much less likely to have some kind of motion blur from the motion of the heart. The final one is the pitch. And so we have a, in the axial direction from the patient's head to feet, there is a certain finite width of that fan beam of x-rays, which has been emitted. So as our x-ray tube is rotating around the patient, the table is moving slowly. And if we define the fact that the table moves one width of the fan beam for the rotation of a tube, we can say that has got a pitch of one. If the table moves twice the width of our fan beam in one rotation of the tube, then we can say that that pitch is two. And effectively, as we're going around, we are essentially leaving gaps. The inverse is possible where we can do some, what we call overscan. And that is where we advance the image and table only half the width of the fan beam per rotation of the CT. And hence we can get overlapping x-rays. And this will give us increased image quality provided that we have got a nice stationary object. Of course, the consequence of the pitch decreasing is that we are moving the table less as the tube is rotating and hence our scan overall will take longer to acquire. Let's look at some of these parameters and their impact on the actual radiation dose to the patient now. The first thing to note is that the tube current times the time is a function of, dictates the total x-rays which are being produced in a given rotation. So we can define a parameter referred to as MAS. So the tube current is the MA and the rotation time is given in seconds. So it's MA multiplied by S. So we get MAS. And so this is effectively determining the x-rays produced per rotation. So if we have a fixed rotation time of one second and we double MAS, then we double the amount of x-rays given out per rotation. Whereas if we have a fixed MA and we halve the rotation time, so we go from a one second rotation to a two second rotation for a fixed MA, then again our MAS is doubled. So our dose or the amount of x-rays in that slice will double and hence our patient radiation dose will double. So the dose is completely directly proportional to the MAS. The tube voltage is a slightly different one. And again, as probably you would expect, as you increase the tube voltage, then the radiation dose to the patient increases, but this is not a linear. And it's approximately a squared relationship. And a good rule of thumb that I always like to go for is that if you increase your tube voltage from a 20 kV step, so maybe we go from 100 to 120 kVP, then this effectively doubles the radiation dose. So what I'd like to bring your attention to here is this curve, these three lines on this graph on the right hand side. And we can see that as the tube voltage is being increased along the x-axis, I've got the dose, which is just a sort of representative dose on this y-axis for three different tube voltages, 80, 100 and 120. And let's just say what I was talking about here, that if we take this 100 MAS point here at 80 kV and we go to 100 kV with 100 MAS, we can see that our dose has gone up from a measurement of two to just over four. So like I said, almost a doubling of dose. It's not a really firm rule, but it's a very good approximation and holds up quite a lot of the time. And we can see, as I was talking about, as we increase the MAS, then the dose to the patient will just increase linearly. Finally, I just want to return to the bullet points on the left hand side. We talked about the pitch. If the pitch is less than one, then remember we're over scanning the patient. So we're going to have greater dose in every rotation, every image slice. If we're increasing the pitch above one, then like I said, we're leaving some slight gaps as we rotate around and hence we're going to get a lower dose per slice. So what can we do to actually try to reduce the radiation dose to the patient? Well, one of the main things in CT is called tube current modulation. And there are two types of tube current modulation that we can do. There is axial modulation and there's rotational modulation. So the axial modulation is something that you would plan or the CT scanner would plan for as it does its planning view. Now, depending on the system, this could be called a scout view or it could be called a topogram. There are other types of names, but they're the two main names that you might hear from it from different scanners. The topogram or the scout view, you will look at the density of the patient as you go down from the head to the foot. And you can see, well, at least some of these points are more dense and some of the points in the middle are less dense. And hence we have this overall shape of the tube current that we expect to be able to need to provide a fairly uniform level of image quality as we go down from head to foot on the patient. Now, there are other types of tube current that we can do as well. And there are other types of modulations that we can do as well. So we can do a lot of modulations on the patient. So that's axial modulation. But in addition to that, there is also something called rotational modulation. And again, what you probably would hopefully can appreciate, this is a nice example on the left-hand side of these two images here, is that if we're going through the shoulders here, we've got a much, much greater distance attenuation of the x-rays in this direction compared with going from front to back of the patient where the distance is much shorter. So if the x-ray tube is over here, for example, then the tube current is going to increase quite a bit more than it would do from when the tube is in front of the patient. I suppose what you can really think about is that the tube current here will decrease compared with when the tube is lateral to the patient. So as the tube is rotating around the patient, then the tube current is going up and down with the rotation. And this will vary depending on the different cross-sectional areas of the patient as we move down through them acquiring the scan. So this is just a nice representation of what's actually happening as we do this acquisition. So the blue represents our, effectively you can think about it as an envelope as we go down from the head to the feet along this x-axis here. And this blue dotted line here is dictated by our axial modulation, which is acquired or planned from the topogram. And then we've got the red, which is the rotational modulation as the tube is rotating around the patient. So now we're going to move on and talk a little bit more about reconstruction. And there are three main settings when we're doing CT reconstruction. We've got the slice thickness, the spacing and the kernel. The slice thickness, very much like voxel dimensions in PET and SPECT, they determine the partial volume effects and in-slice noise. If you get smaller slices or thinner slices, your transaxial images are going to be noisier, just like if you're going from a big voxel in SPECT or PET to a smaller voxel, you've got less counts in every individual voxel and hence it gets noisier. So you would compensate for that by increasing the tube MAS. The spacing can be used to determine the overlap. And this actually is a nice way of making the images look a little bit more cosmetically pleasing, if you're looking at them in the coronal and sagittal views, because if you have slice overlap, then you have a much more smooth transition as you go down structures through the body. The final thing is the kernel. And again, the kernel you can almost think about as a bit like a smoothing filter in emission tomography PET and SPECT. And the kernel is typically defined by, depending on the type of system that you're using, a number like we see here. And typically the higher the number, the sharper the image and the more noisy it is. So these are two images from our PET CT system that have a pre-constructed with a B05 kernel and a B32 kernel. And this 32 kernel is a pretty typical reconstruction and you can see the nice sharpness. This B05 kernel is a very, very smooth reconstruction and is primarily intended to be used for attenuation correction. So now we're gonna talk a little bit more in detail about quantifying the patient radiation dose from a CT scan. And the three main quantities that we are interested in. The first is the CT dose index, okay? And by the unit volume. So this is effectively the average radiation dose in a standardized unit volume. And it's actually defined within a perspex volume in a phantom. And this is then for a standard set of reconstruction parameters and acquisition parameters, this is then modified for the patient and it knows from how much attenuation is happening in the patient, what typical CTDR would be. And this has got units of milligray. So like I said, this is just per centimeter or per unit volume. Then we have the dose length product and the dose length product is the average dose over the total axle scan length of the patient. And it's effectively a product of the average CTDI because remember the CTDI will actually change slice by slice according to the dose modulation. So over the entire scan length, it will have an average value. So this dose length product is that product of the average CTDI over the length in centimeters. And that will have units of milligray centimeter. So these are essentially image derived or scan derived parameters that don't have a great deal of meaning until you do something with them to actually translate them into a patient effective radiation dose, which we are used to hearing about in PET and SPECT when we're measuring, looking at things in sieverts. So when we've got our DLP, we then need to look at what region of the body we've acquired the data over. And we've got some conversion factors which you have for that. Now, one of the important things that these conversion factors are really only defined for a standard person or a standard patient. So a 75 kilogram patient. The translation of that into a much larger patient is actually quite thought with quite a degree of uncertainty. So really when we're calculating radiation dose from a CT scan, it's only ever really an approximation. So that is one word of warning that I would try to get across here. But like I said, these have units of millisieverts is which what we're used to hearing about when we're looking at other radiation exposures. So we've got this process mapping here of how we take our scan parameters which are on the left-hand side here, our KV, our MAS and our pitch. They feed in and these will determine our CTDI volume. Okay, remember that's the dose per unit volume. When we've got that CTDI volt, we can then feed in the scan range and that gives us our DLP. And then finally, once we've got our DLP, we feed in the body part that we're looking at and this gives us our patient radiation dose estimation in millisieverts. So the one thing that I haven't really talked about is these conversion factors. And these are the conversion factors which are used. And as you can see, these are the effective dose in millisieverts per milligray centimeter. So the milligray centimeter here, that's our DLP number. So to work out the millisieverts to the patient, we take our DLP and we multiply it by one of these conversion factors depending on the area that we're looking at. So in the realm of nuclear cardiology hybrid imaging, we're typically going to be looking at the factor of 0.017 over the chest, over the thorax and the heart. So these are just some example parameters. Obviously disclaimer is that these are what we use clinically on our specs and our PET systems and yours may differ entirely depending on what type of equipment that you have. But nonetheless, it gives you an indication of what we should be thinking about when we're doing CT for attenuation correction only. And here we've got our DLPs for the spec and the PET and our typical radiation dose, which you should see. So I think the point to get across here is that these are very low doses. So well under one millisievert for CT attenuation correction. So speaking of attenuation and correction, this is the next part that we're going to talk about. So when we have our CT image, we use this to generate an attenuation map. And the attenuation map is a spatial distribution of the attenuation coefficients for the given radionuclide energy that we're interested in acquiring the data over. So this graph here shows how we do this. We do it with a lookup graph or a lookup table effectively. So we take our Housfield units from our CT on the X-axis and then we find how it corresponds to a lookup process. So I've shown this for four different energies here. For spec, we'd typically be interested in the 140 keV for technetium. And for PET, we're looking at 511 keV. And from here, we take a CT and we find out what the corresponding attenuation coefficient would be. And these will be the units that we see in this attenuation map here. Now, again, what I want to get across is that this conversion process is extremely robust. And so, as I've said previously, as we've seen previously, the CT dose for attenuation correction can be extremely low and really should be extremely low, easily below one millisievert. So let's look at how we actually calculate attenuation when we've got this attenuation map. So we take a, what we have to imagine here is that we've got a PET or a SPECT image overlaid onto this. And let's just take a voxel here and we'll say, okay, well, some gamma rays are being emitted from here. So what we will do is we will calculate and trace a ray path from the origin voxel all the way through the attenuation map to the edge of the patient. And along this path length, what we're doing is we're calculating the total attenuation which has occurred through all of these different densities of voxels through the attenuation map. And then we can use that to correct the attenuation that we would expect to see from that gamma ray source, whether it be SPECT or PET. So one thing that hopefully you can appreciate is this attenuation map has got to include the entire patient. So this boundary, because remember, we're tracing a path from an emission point to the edge of the patient. So if we don't measure the edge of the patient accurately on this attenuation map, then we're going to run into difficulties. But there is one exception to this rule here. And this is when we're looking at SPECT. So this is a CT and an attenuation map as if we are looking from the patient's feet, as is convention when we're looking at x-ray images. So our SPECT orbit on a traditional gamma camera acquiring in 90 degree configuration is to perform an arc from right anterior oblique to left posterior oblique. And if the heart is here, then we're going to trace the path of these gamma rays from the heart to the positions that we're acquiring the data on, like so. And what you'll notice here is that the attenuation map is a bit bigger than the CT. And that's because quite often in SPECT, we'll apply some zoom to the acquisition. And depending on the type of system that you're using, then you might also apply the zoom to the attenuation map. And as we can see, if I remove my arrows from this, if I remove my arrows from this slide, you'll see that we've chopped off slightly the edge of the patient. So at some angles, that tracing of that ray path is not going to be correct. So this is something to be aware of, and this is called a truncation artifact that can occur. Now, what you'll notice if you're being keen-eyed here is that the right-hand side of the patient we're not interested in at all. So we never trace any gamma ray paths in this direction because we never place a gamma camera on this right-hand side of the patient. And so if we shifted the patient in this case here over to their right, then we probably would have not had any of that truncation occur. Now, in PET, we are interested in tracing the entirety of the line of response through the patient. So this dash, this solid line here through the patient is what we're interested in measuring the attenuation, and these dash lines will show their path on the way out to the detectors. So again, all of the patient now needs to be within the attenuation map. So if you've got a very, very broad path, so if you've got a very, very big patient that's truncated, then again, your attenuation correction is not gonna be completely accurate. However, one of the things that has been acknowledged by manufacturers is that this does occur, and there are ways of getting around this. So a standardized CT field of view is in the transaxle direction, is 50 centimeters. And you can see if we've got a large patient here, and this is a case of a patient with their arms down, you can get these bits here where they're missing. So if you were to take this CT image and convert it into an attenuation map, you'd have some portions missing on the attenuation map, and this would be wrong. So what you can now get on quite a lot of systems is what has been referred to as an extended field of view CT reconstruction. And this takes some of the information from the partial views as it's rotating around to fill in the blanks effectively. And so we can get a nice estimation of what the actual patient is outside the standard 50 centimeter field of view, which you can see from this red dashed line here. And so now we've got a full patient here, and if we convert this to an attenuation map, we've got all of this in the patient. So finally, I'm just gonna finish up by talking about calcium scoring. So this was established in the 1990s, and one of the main measurements is called the Gadsden score. Now it's worth knowing that this was actually not really defined on a conventional multi-slice CT that we know and use in hybrid imaging. It was actually done on an electron beam CT with extremely fast rotation, faster than you typically see on all of your, even on your most modern multi-slice CT systems. It was done on three mil slicing on ECG triggering, and actually it was done on breath hold CT, which again is quite rare to do for attenuation correction. But again, it's something that you would want to do for calcium scoring. The actual score itself is based on the area of calcium, which is identified over a given threshold. And we use that threshold of 130 HV units, and there is a density factor according to the maximum HV units within a given region of calcification, which has been identified. Now, interestingly, there was no mention in this paper of the reconstruction method. And again, there have been some works out there which have shown the impact of image reconstruction on different calcium scores. But it can now be performed routinely on a multi-slice CT, multi-detector CT. And one of the things that we know is that like all small objects, if you've got motion, then if you just do a non-gated CT or an ungated CT, then you might end up blurring some of that calcification out. In addition to that, if you're looking at the Gadsden score, remember it uses that one maximum CT value in an area. So if that's being blurred, that maximum HV unit may be reduced and it may fall through a different density factor scaling and hence could have a significant impact on the calcium scoring. Having said that, high calcium can very easily be visualized on even a low-dose non-breath hold, non-gated attenuation correction CT. And this will certainly provide you with useful information. And in fact, now there have been some studies being performed where there's been a comparison of calcium scoring on low-dose ungated CT with gated CT. And these results are quite promising. And again, looking at it with the effect of image reconstruction, there's been some work done here. And this was shown to be relatively robust. But again, like I said, one of the biggest impacts is actually the kernel. So if we compare this work here, then if we take these two protocols, A and B, you'll notice that the differences are quite small. And if we look, all that's been happening here is that this last thing, this has been changed. However, if we take protocol A versus protocol G, so protocol A was the reference, we'll notice here that a different kernel has been used and there was an enormous change, which is not immediately apparent, but it's only when you actually take note of the vertical scale on these two graphs about how much difference really is in these images. So it's very important that when you're setting up your parameters for calcium scoring, you don't change the kernel. So to summarize, modern multi-detector CT scanners are available for SPECT and PET, and these will allow for very low-dose CT for attenuation correction. There are tools now available to minimize truncation, and the calcium scoring can be performed routinely. Both of these systems will allow for ECG-gated acquisition to do a proper calcium score. So thank you very much for listening, and I hope that was informative and interesting.

CT physics

nuclear cardiology

x-rays

radiation dose

attenuation correction

calcium scoring

image reconstruction