Transcription
Hello everybody and welcome back. Today, we're going to be looking at the physical process of acquiring a CT. Specifically, we're going to examine the relationship between how the X-ray source rotates around the patient and how the patient moves through the CT machine during data acquisition.
And depending on how we set up this acquisition sequence, we can separate acquisition into two separate modes. The first mode that we're going to look at is an older mode that we generally don't use anymore, known as axial or sequential CT acquisition. The second mode is called helical or spiral CT acquisition, that we use in most modern CT scanners today.
So let's start by looking at axial acquisition. The patient is lying within the CT machine, and our X-ray source is going to rotate a full 360 degrees around the patient. The width of the beam at the isocenter is going to be the usable data that we can use for that specific section. Importantly, in axial or sequential CT acquisition, at this point, the X-ray source is turned off. We then move the patient the exact same distance as the width of the beam at the isocenter. Here, so that happens here, the X-ray source is off, and we move the patient. We then repeat this process. The X-ray source rotates a full 360 degrees around the patient. We use all of the data from that 360 degrees to generate the data that we're going to use to process and ultimately display our image. Again, we turn the X-ray source off and rotate another 360 degrees around the patient. Each time the patient is moving, we're only moving them the width of the beam at the isocenter. We continue this process until we've covered the full area of anatomy that we're interested in. This is what's known as the coverage of the CT machine. In this example, our coverage is five times the width of the beam at the isocenter.
Now, we haven't created five slices. You see, the CT machine that we're using here is a multi-detector CT, an MDCT, that has eight separate rows in this example. So, in this example, theoretically, we can generate 40 different slices for each beam that we acquired. We can generate eight different slices.
Now, the major issue with axial or sequential CT acquisition is that time when we have to pause the CT machine and move the patient. Now, this was a function of the hardware of the CT machine itself. We couldn't continually rotate the CT machine because the wiring that was supplying power to both the detectors and to the X-ray source would get wound up after one rotation. We have to stop the CT machine, unwind the wiring, and then repeat the process again. And all the problems with axial or sequential CT acquisition generally boil down to this: that pause means there's a period of time when we're not generating any usable data, and it means it's going to take much longer to acquire the image that we want to acquire. Ideally, if we were imaging the thorax, like we were looking at in this example, we would be able to generate an image represented by this diagram here without interruptions between each of those pausing of the X-ray machine. In practice, what often happens is we get misregistration of that anatomy, mainly because of patient movement during that pausing period. If the CT scan takes too long to acquire, the patient may not be able to hold their breath for that long a time. The chest cavity may expand or contract, and this involuntary movement that we can't control, say bowel movement or the pulsatility of, say, the aorta in the thoracic cavity. So the issue here is pausing and stopping that X-ray source and then moving the patient. Movement can happen in that time, and it takes much longer to acquire.
Now, we've mentioned it before that the development of slip ring technology meant that we could continuously rotate the X-ray source around the patient without the need to pause it and unwind those wires that were supplying the X-ray source. Doing this means we can continually expose the patient to X-rays, and whilst we are exposing the patient to X-rays, we can move them through the CT machine without needing to pause that X-ray source. And that's the foundation for helical or spiral CT acquisition. Now, you'll see those terms used interchangeably. Helical or spiral. A helix is a type of spiral. I prefer to use the term helical because it describes a spiral that is keeping the same distance or the same diameter of its circle. A spiral could get smaller in diameter or could get larger in diameter, but it doesn't matter what you use. Both terms are acceptable in CT imaging.
Now, have a look what happens here. As the patient moves through a rotating X-ray source, we generate data. And you'll notice that that data is no longer parallel to the X-ray source exactly, and it's no longer perfectly perpendicular to the patient. We get slating of some of that data here. That's because the patient is moving whilst we're acquiring the image. Now, the multiple variables that we can change here that's going to change this geometry. For example, if we take this patient where the X-ray source is rotating at exactly the same speed as in the previous example, but we're moving the patient much faster through the CT machine, you see how data acquisition is going to change. Now, what we've done here is we've just sped up the patient by a factor of two. We've moved them through twice as fast as before. Now, it turns out that the relationship between how fast the patient moves through the CT scanner and how quickly we can rotate the X-ray source around the patient, as well as our beam width, we can describe that relationship with a value that's known as pitch. And pitch becomes incredibly important in helical or spiral CT acquisition. Pitch describes how fast the patient is moving through the gantry of the CT machine, that's known as a table speed, in relationship to how quickly we're rotating the source around the patient. So the faster we move the patient through the CT machine, the faster the table speed is, the higher the pitch is going to be. This example here is a higher pitch than this example here. The longer the rotation time, the higher the pitch is going to be. The longer it takes to move the X-ray source around the patient, the more likely we're going to get gaps in our data acquisition. That's associated with a higher pitch. The narrower our beam, the less coverage we get in the Z-axis direction. So narrower beams are going to give a higher pitch. The wider that beam, the more coverage we get, more overlap of data that we're going to get.
Now, when we have a pitch of one, the patient moves through the CT scanner, the distance that they move will equal the width of the beam at the isocenter. It means for a full 360-degree rotation of the X-ray source around the patient, that patient would have moved through the beam at exactly that width of the isocenter. We would have covered, we'd have no gaps within our data. Any pitch higher than one, we're going to start to get gaps within our data. Any pitch lower than one, we're going to get overlapping data, and we're going to increase the amount of dose that our patient gets.
Now, there's another way to represent this data acquisition. If we look at our X-ray source rotating once around our patient, the region of anatomy that's going to be covered fits in this shape here. It's kind of a bizarre shape. We've got a 360-degree rotation, but there's offset in the Z-axis, and that offset is represented here. This is not a perfect circle around the patient because of that movement of the patient through this scanning machine. As we continue this process, we are going to acquire 360 degrees of data for sequential Z-axis slices through our patient, but those slices aren't like the axial CT scanning where we got perfectly perpendicular slices. These slices are now offset. What we can do is take the data that we've generated for all 360 degrees and we could send that data and create sinograms for each one of these 360 degrees. But when we then take that sinogram and process that data and create the image, that image is going to represent anatomy that's slightly off in the Z-axis direction.
Now, there's a way that we can get around this. We can move that data together, squish that data, and we can process that data in a process that's known as interpolation, and that becomes incredibly important in helical or spiral CT acquisition. Let's look at how we were representing this data previously when the CT machine was going around our patient. This data representation is the same as this data here, but we're looking at the patient from exactly side on. If we zoom into this and say we were interested in the anatomy at this axial slice here that related to a detector at this location. If we were to use the same detector after 180-degree rotation, that detector would now be in this location because the patient has moved slightly in the Z-axis. What instead if we looked at the detector that was exactly perpendicular to our initial detector after 180-degree rotation? Once that X-ray source has moved around 180 degrees, we're using a different detector that's further over on the X-axis here, but lines up initially with our initial slice that we're interested in. What we do here for that full 360-degree rotation, if we can move that detector around to correspond to an exact axial slice across our data set, we can essentially slice this data set and create a perfect axial slice. Moving which detector we are using to generate our sinogram for this axial location in the Z-axis is what's known as interpolation. If you think of this as a cake, now we can slice it perfectly down a single slice. We account for this offset that was caused by the patient moving through the CT machine.
Now, that's all good and well when we're covering all of the anatomy. There's no gaps in our cake here. What happens if we're using a higher pitch where we've got gaps? This is where interpolation really comes into its own. See here, we've got continuous data. So the data that we sample in that 360-degree ring is going to give us the actual anatomy that's lying in that axial slice. With a higher pitch, we've got gaps in that anatomy. Now, you may have heard of extrapolation, where you take a sequence of data and you say, well, I've seen what the sequence is looking like, let me extrapolate what that sequence I think would look like. That's extrapolation. Interpolation is getting a sequence of data, but there's gaps in that data, and you're trying to figure out what makes the most sense to fill in those gaps in the sequence. So here, our detector, when it's facing this way, is saying measuring a value of five. Now, this is an incredibly simplified version because obviously there's multiple attenuation values across that entire row. Let's pretend it's just creating one value of five. There's missing data for at least a full 180-degree rotation on this side of the patient. Now, what we can do, in basic terms, this is just a good way to think about it, is the machine can say, well, what was the value immediately to the right and what was the value immediately to the left? And let's average those out. So say this is probably going to be seven here, between these two values, and I predicted five on the other side, the opposite side of this scan. So maybe let's go and fill in this data as being the six, lying between five and seven here. And that's the concept behind interpolation. Obviously, it's a lot more mathematically complicated here, but that's a good way to think about it. We can fill in those missing gaps, but it's not going to represent the true anatomy here like we had with a pitch of one.
Then why would we want a pitch that's much higher? Well, I've kind of touched on it before. A higher pitch is going to equal a lower dose. Pitch is proportional to dose. It's inversely proportional to dose. The higher the pitch, the less dose for a specific region of anatomy that we're going to have incident onto our patient. Obviously, CT scanners use X-ray radiation, ionizing radiation. It's not innocuous radiation, so it's a trade-off here between how accurate our image is going to be and how much dose we want to expose the patient to, and sometimes how quickly we need to acquire the image. Higher pitches often allow us to take the image much quicker because we cover more anatomy quicker by function of moving the patient through this helical machine faster. So that's the first issue that comes with helical scanning is that we need to interpolate the data in order to get accurate axial slices.
The second issue, you might notice that on the lateral borders of our coverage here, we've got regions which don't have a 180-degree counterpart. We can't interpolate the data here in the first 180-degree turn or the last 180-degree turn in helical acquisition. This wasn't a problem in axial or sequential imaging because that 360-degree rotation gave us all our anatomy perpendicular to the patient. Here we have that problem because now our data is slanted like this. What we can do is place a collimator here that prevents X-rays from reaching that first half, that first 180-degree rotation. None of that data here was going to contribute to our final image. Now, this collimator is what's known as an adaptive beam collimator, specific to helical or spiral CT acquisitions. It can either be within our CT machine here, within the collimators that move as that CT machine rotates around, or it can be within the X-ray machine itself as a fixed source that prevents X-rays from being incident on the patient in this part of the scan. Obviously, the same can be done at the end of our scan here. And you may be wondering, is that really going to make much of a difference to dose? Well, if we were taking an image say that required 10 rotations around a patient, the first half rotation and the last half rotation aren't going to provide us any useful data. If we were to use adaptive beam collimation to prevent those half rotations from being incident on the patient, we would have the nine full rotations that are generating valuable data that we can actually create our image. We've reduced the dose in this example by 10%. We've no longer got 10 full rotations, we've got nine full rotations. And depending on how big our coverage is, for relatively small coverages as a percentage, this adaptive beam collimation is going to be a fairly large percentage. If we have a very wide beam, as well as a percentage, we're going to have a large percentage reduction in dose. A wide beam, if we're cutting out half of that beam at the start and the end, we're cutting out a lot of radiation to the patient. So interpolation and adaptive beam collimation go hand in hand with helical or spiral scanning.
Now, there's many ways that these kind of concepts can come up in exams, especially when we're looking at things like pitch and we're looking at things like dose. I've seen questions asked where they say, if I was to change the cone angle of our beam, how would that influence pitch? Now, we've looked at X-ray beam geometry. We know that cone angle is going to increase the width of the X-ray beam in the Z-axis. It's going to increase the width of the beam at the isocenter, and an increase in width is going to lower our pitch. So knowing that relationship between the X-ray beam geometry, the patient moving through the machine, the X-ray source rotating around the patient, really comes in handy when it comes to answering questions in exams.
Now, we're going to move on to the exciting parts of CT. We've got most of the boring parts out of the way now. We're going to look at the actual generation of our images. First, we're going to look at Hounsfield units and how they represent the attenuation data from within our patient, and then we're going to look at how we go about generating those Hounsfield units using image processing, back projection, filtered back projection, iterative reconstruction. All exciting topics that are a little bit difficult to understand. So hopefully, we can really solidify those concepts. So until then, I'll see you all. Goodbye.