 Good morning. I'd like to thank the organizers for inviting me to give this talk intro to 3DTE. Here are my disclosures. The objectives of my talk are first to review an approach to 3D echo data acquisition, and then I'm going to also discuss some common issues with 3D echo data set acquisitions and solutions for them. I'm going to use a combination of trans thoracic and transophageal echocardiographic imaging as the concepts for acquisition are the same regardless what majority you're going to use. The other thing is that the vendors will have different nomenclature. As long as you understand the terminology and what that means, you'll be able to optimize the images on your machine no matter what vendor you're using. Whether there's a significant difference in nomenclature, I'll try and point it out on my slides. I'd like to also direct the audience's attention to the guidelines that were published in 2012. There are new guidelines coming. They'll actually probably be broken down into two parts, one on valvular acquisition and analysis, and the other on cardiac chambers. So this is sort of the outline of what my approach to a 3D data acquisition is. I'm going to discuss each of these steps in detail, but overall the first step is to optimize your image and then to make decisions on acquisition modes. First, what size of pyramid that you want, and then second of all, depending on the size you've chosen, whether or not you need to go beyond single beat to multi-beat acquisition and then whether or not you want to add color doppler. Now your choices for acquisition mode are going to be driven by your need for spatial versus temporal resolution and whether or not you have the ability to deal with gating artifacts. Then I'm going to discuss rendering options, whether or not you want to crop your data set and whether or not you want to adjust your thresholds, and then finally I'm going to talk about how to display your image and whether or not you need to do any further analysis using software packages. So the first step in acquiring a 3D data set is to optimize your 2D image quality. If you look at the image on the left, you can see I've used multi-planar mode, so you can see the cross planes of this mitral valve as well as the 3D data set. And you see on the multi-plane those that in the left atrium, there's actually some artifact that is causing some noise in the left atrium when you look at your 3D data set. Now if you look on the image on the right, we've actually found a different angle where we have a much cleaner left atrium, and you can see that on that picture, the mitral valve is actually visualized to much nicer and you don't have any problems with the noise anymore. So once again, if you have poor 2D images, you'll get poor 3D images. So you really have to spend the time to optimize your 2D image before you acquire your 3D. The second thing you need to do is you need to use multi-planar mode when you're acquiring your 3D. So here you can see that in this view, we're trying left ventricle. The top left panel is the view on your 2D image that you've used before you got into your 3D, that you're going to use for your 3D acquisition. The top right and the bottom left images are cross planes of that 4-chamber view that you were using. What you want to do is in all of these three views, your original 4-chamber view, your cross-plane in 2 and your short axis, is make sure the structure that you're interested in is in the pyramid throughout the cardiac cycle, insisting in data sleep, and then you also want to make sure that the definition of the structure is very clean during the cardiac cycle. And that way, once you've acquired your data set, you'll know that you have a good image and also, depending on how you want to cut through your image, your 3D data set, you'll be able to see everything that you're interested in. One of the things you can do to improve your image quality is try to image the structure perpendicular to your being. So this is why on trans-sophageal echocardiography, your mitral valve looks really nice because it's perpendicular to your ultrasound beam. And this is also why your aortic valve on the midisophageal view tends to not be as nice, and you deal with leaflet dropout, especially if the leaflet is thin and in a healthy person. Now, once you've optimized your image, then the next step is to decide what pyramid size that you want. What you want is you want a pyramid size that covers your object of interest and then you want it to be a little bit bigger than that to give you some landmarks around it so you can orient yourself. Now, in the early days of 3D, there were only certain pyramid size that you could use. They're relatively fixed. And then you can only do either single bead acquisition or multi bead acquisition with them. And this day and age, you can use almost any pyramid size with single or multi bead. And you can also do color Doppler acquisitions with either single or multi bead. Now, when you look at the machine, they're actually going to have different presets for the pyramid size. Now, these presets are more like starting points on current machines because you can increase or decrease the size of your preset, depending on what your object of interest size is. And so usually you want to use the zoom or the smallest size when you're looking at small, fast moving structures such as valves or masses. But remember, you've got to make sure that it's big enough to include some of these adjacent structures so you can orient yourself. I tend to use narrow sector or live if I'm in a procedure and I quickly have to get a larger area and including the catheters as well as the valve that I'm guiding the procedure for. And then I use wide sector mainly if I'm acquiring a left ventricle or the whole heart and I need a big data set there. This is a slide sort of telling you sort of the relationship between the different vendors and the terminology on their knobs in terms of the size of the pyramid that you're going to start with. Now, why does pyramid size matter? Well, it matters for two reasons. First, if you're actually doing volumetric measurements, you want to make sure that you capture both the largest and those the smallest volumes in the data set to create an accurate ejection fraction. So most vendors gate using ECG and they use the QRS complex. So that means N-dialy is usually captured on the 3D data set. However, if you have a low volume rate, then you may miss n-systole. If you miss n-systole, you're going to overestimate the n-systolic volumes and then underestimate the ejection fraction. The other thing is if you have a small moving fast moving structure, if you have a low volume rate data set, you may not actually see it in your 3D data set. And so that's why you want to get a higher volume rate to make sure you don't miss these fast moving structures, whether it be a vegetation or a mass. Now here I have two examples of a left ventricle and one is acquired, both are acquired using four-beat acquisition. The one on the left, you can see, is a smaller pyramid than the one on the right. The one on the left only shows you very well the left ventricle and left atrium. The one on the right includes the entire heart. So you can see the left ventricle as well as the left and right atria. And you see that on the left one, the smaller pyramid, you get a higher volume rate of 30 Hz versus the right you only get 20 Hz. When you look at the walls of the left ventricle as it comes in, you can see the one on the left with the higher volume rate. The walls come in much more smoothly than the one on the left. And that's because the one on the left is not sampling enough so that you can see now if you're only at 20 or even 15 Hz, how you could miss n-cisly and so get a wrong attraction fraction. Now I've alluded to this before, but there's a trade-off in terms of your relationship between your volume rate, pyramid size and image resolution. So if you have a certain structure, if you look at the panel on the left and it's covered by certain pyramid size, you get a sort of certain volume rate as well as an image resolution. If you increase that pyramid size to cover something larger and you want to maintain your volume rate, then you'll sacrifice image resolution if you look at the panel in the middle. To try and recover that image resolution, you'll have to decrease your pyramid size a little bit while you're maintaining your volume rate and that will improve your resolution. However, what if you have a structure that has to be covered within a certain pyramid size? You need a certain volume rate and image resolution. Then what do you do? This is where you need to move some single beat to multi-beat acquisition. Now, what is single beat acquisition? So single beat acquisition is the acquisition of multiple pyramidal datasets covering your structure of interest per second in a single heartbeat. Multi-beat acquisition, you are covering several small sub-pyramids that only cover a portion of your structure and you're then going to fuse them all to create your larger image that shows the entire structure that you're looking at. With high volume rate mode, what you're doing is you're actually it's an interpolation algorithm. So the probe, all the elements in the probe are not active and scanning at the same time, which is how you can get a faster volume rate. However, the blocks that are not being scanned, the machine comes up with an interpolation or a guess on what should be in that make the image look complete. And then in the second cardiac cycle, it tries to fill in that. And so that's why because it's guessing on some images and it's filling in with real pictures on others, that's why the images are not as nice as a multi-beat acquisition. So here I'm showing you a difference between the single beat versus multi-beat and high volume rate modes. So this is the left ventricle. If you look across the pyramid size, it's the same for all of these acquisitions, except the only difference is going to be where I use one beat on the left, four beats in the middle and high volume rate on the right. You can see on the single beat acquisition on the left, it's only six hertz. And so the walls of that ventricle as they come in, it's very choppy. It's not very smooth. When you go to multi-beat in the middle, you can see that the walls are coming in much more smoothly and you actually have a much nicer picture and you're getting a much higher volume rate at 23 hertz. Now, if you look at the high volume rate mode, you can see once again, you get a high volume rate of 23 hertz compared to the original single beat. But the image quality is not as nice as that of the four beat. Depending on what vendor you have, you'll have the different options for your acquisition until you just have to know what vendor you're working with in order to know what you can do to acquire a higher volume rate image. Now, one of the things I do sometimes, if I don't have the time to acquire or I don't have the ability to acquire a multi-beat acquisition. So here, you can see the ECG. You can see there's a lot of significant arrhythmia in this patient. So the multi-beat is going to end up causing me a lot of artifact and trouble. So I'm using a narrow pyramid and I'm actually sweeping. So this is a prostatic mitral valve. There's a catheter going through the valve instead of a paravalve leak that we're hoping for. And I'm scanning back and forth, actually, showing where the catheter is and its relationship to the interatrial septum. Now, why don't we use multi-beat all the time if it can give us much better pictures? Well, the problem is when you're fusing all of those sub-pyramids, is that you can end up with stitch artifact if the fusion is not done properly. So here I've got a couple of examples. If you look at the bottom left, this is a mechanical aortic valve and you see there's a stitch going right through the middle of the data set on the bottom right. You can see there's a mitral valve color data set and you see there's a big stitch artifact going through the valve there. And then on the top right, you can see this is the left ventricle. It's a little bit more subtle on this data set, but during cis-league, you can see there's a line going through the middle of the ventricle and there's the stitch there. So how do you address stitch if you actually need to use multi-beat acquisition? Well, first of all, acquire using this multi-view plane. So if you look at the color Doppler data set on the top left, it's a mitral valve and you can see that on the cross planes, you can see that they're black before the data set comes in. So we've acquired this data set too soon. We haven't waited for all the sub-pyramids to come in and then we haven't waited for them to fuse properly before acquiring. And so you won't see the cross planes coming in on your main image. You have to use the perpendicular planes to your sweet plane to actually see this. So this is why multi-plane is very valuable. The second thing is you have to see why the cause of your having these stitch artifact. Is it because the patient's moving? If so, try and stabilize the probe and ask the patient to breath hold. If they're ventilated, try holding the respirator to see if you can get your multi-beat data set before ventilating the patient again. And sometimes you can't do anything and you'll just have to use other single-liter high volume rate mode to try and get your image. Now if it's because of an arrhythmia, sometimes you can find a spot where the arrhythmia is more quiet during the acquisition and then you can use a retrospective acquisition setting where you freeze your cine plate and then you trim your data set to acquire the fused beat. Or you once again, you may have no choice but have to use single-liter high volume rate acquisition to get your data set. Now everything I've described for non-color 3D data sets also applies to color 3D data sets. The one thing you have to be aware of even though there has been progress in this area is that with 3D color data sets, you still have a drop in volume rate. And so sometimes you may have to use a smaller data set when you go into the color part, like a smaller pyramid, or you have to go into multi-beat. So once you've decided about the mode of acquisition that you're gonna use, you have to decide whether or not you want to crop your data set. If you crop before you acquire your data set, all the information you've cropped out cannot be regained after you've acquired your data set. If you crop after, all that information is still there. However, the problem is then you've acquired a larger data set than you really needed, and you so may end up with a smaller temporal or spatial resolution that you, or smaller, sorry, temporal resolution, than you wanted. Okay, so you'll have to make a decision about that before you crop. Now, there are a couple of different methods to crop your data set. If your object of interest is aligned perfectly within your data set, then tools such as AutoCrop, which automatically cut your 3D data set in half into the middle of your pyramid, wherever it may be, are useful. EmbarkCrop, which crops along the x, y, and z axis through your data set are good to use. But if your data set's acquired a little bit off axis or the structure you're interested in is actually not aligned perfectly perpendicular or parallel to your x, y, and z planes of your 3D volume, then you may use to use something like a plane crop or a view crop, which gives you that flexibility of tilting over a little bit and cutting in these off-axis planes. So sometimes the mitral valve may be aligned properly, but the tricusp valve is a little bit crooked, and so this is where that comes handy if you want to look at tricusp valve anatomy. Now the tool that I use a lot that's actually very handy is these quick crops or that let you just crop on either side of your object of interest, and then everything outside of those two lines is excluded. And then the structure that you're interested in usually for me, it's a valve, is just there available for you to actually look at. Now there are newer tools available on the machines. This let you quickly click on the 3D data set and then your cross planes are seen. Mostly the vendors also have these options for automated view cropping. You take a very large data set where you cover everything that's in the heart and then you have to do some sort of initialization where, so that way they know where the structure is located. And then you can just hit like a four chamber view and it'll crop automatically to the four chamber view or the mitral valve. So once you've cropped your data set and you can see the structure that you're interested in, then you have to render your data set. Now what does rendering mean? So if you think about it, you're taking a 3D structure and then you're creating a 3D data set. But most of the ways we display our 3D data set is actually on a 2D video screen. And so what the rendering does is actually make this 2D image look more 3D like. And the main three things that you're going to adjust are usually going to be your gain, your smoothing, and your brightness. Now there are other modes that are available on these rendering settings that can make things look more like tissue light or like there's a light behind the data set to improve your visual license. And all they really are is just trying to make that data set look more 3D like. Now gain is the main one that you're going to be adjusting a lot of the time. One of these you have to realize that if you over gain, you end up with a lot of noise or snow on your image. If you under gain though, you're going to have drop out of your tissue and then you won't know if that's truly, is there perforation? Is there a gap there? Is there a cleft? Or is that just because you've dropped your gain too much? So you have to be careful when you're optimizing your image that you don't do too much or too little of your gain. Now there are two ways of adjusting your gain. First is overall gain. So this changes your gain everywhere in your data set evenly. Then there's the regional gain. Regional gain is usually only adjustable before you acquire your data set. So you have to adjust it when you're setting up your image. I usually, for regional gain, I'm playing with my TTCs and I'm decreasing the gain in the left atrium when I'm looking at the mitral valve and increasing it as a level where the leaflet tissues are. The other thing I adjust a lot is smoothing. So smoothing is if you, lets you see the fine details of the structure you're interested in. You don't want to decrease it too much. So if you look at the image on the left, we've decreased a little bit and you can see the sort of the structure of the amplaster device so you look at the top left. However, the problem is if you decrease it too, sorry, increase it too much, then you end up with too much, or sorry, you decrease it too much. You end up with too much granularity and it's too much noise on that picture. If you increase it too much, you smooth out all the details and so you can miss a lot of the finer structures that may be there that you want to see if you look at the image on the right. And then brightness is the last rendering tool that you should adjust to try and improve your image if you are having issues. Now brightness, you kind of think of as a flashlight. If you make it too bright, then it whites out all your structures and you lose that 3D depth. If you make it too dark, once again, it's too dark and you lose that 3D depth. And so you want to increase it to just improve that 3D ability or 3D-ness of your picture. Now finally, color. So a lot of the times when you acquire your 3D data set, the color is everywhere. One of the quick tools that I actually like to do is I change the color filter. Now when I increase the color filter, this lets me see the mange up but I lose the smaller jets in the picture so you have to be aware of that. But it lets me quickly see where I am in the middle of a procedure and where the jets are coming from so I can optimize that quickly. Once you've rendered your data set, then you have to describe how you want to display it. The guidelines actually provide recommendations so how you should orient your images as well as your chambers. Your display, most of the time now we've talked about volume rendering for the display if you look at the top right. However, you can actually analyze your data set and create models, either be a mitral valve model or a left ventricle model. And then either present a surf rendering or wireframe rendering of that model. The other thing is you can do slices through your 3D data set and present it as such. Now there is a lot of work being done to actually let these 3D data sets be presented in a 3D manner. Some tools that are being used now include stereo glasses or stereo monitors to let you see the 3D shape. There is work being done with holographic displays. One of the things that is actually happening for a lot of pre-planning for procedures is the 3D printing mode. And then very popular right now is development of virtual reality or augmented reality tools. So in summary, these are the steps that I would recommend for acquiring a 3D data set. First, you want to optimize your image. Then you have to make a decision on your acquisition mode. The first thing is the pyramid size. And then once you've decided upon your pyramid size, is the volume rate high enough? Do you have to go to multi-beat? If you have to go to multi-beat, can you deal with the gating artifacts? And then whether or not you want color or no color on your image. After you've decided your acquisition mode, then you want to decide about cropping, whether or not you want to crop before or after, and then how you want to crop. And then you want to adjust your rendering thresholds. Do you want to adjust your gain, your brightness or smoothing, or perhaps change the characteristics of the tissue. And then finally, you want to decide how you want to display it. And then whether or not you need to do any further analysis to get more quantitative measurements from your data set. Thank you.