Three-Dimensional Ultrasound in the Abdomen: How, Why and How - SD

Introduction to 3D Ultrasound in the Abdomen

Hi, I'm Dr. Franklin Tesler, chief of Body Imaging at the University of Alabama in Birmingham, Alabama.

I'm going to be talking to you about 3D ultrasound in the abdomen to give you some ideas about where 3D ultrasound will help you with abdominal patients.

I'm going to be talking about three dimensional ultrasound in the abdomen.

In the next 30 minutes, I'm going to be talking to you about three dimensional ultrasound, beginning a little bit with instrumentation, how it's done.

Then switching to the clinical applications in the abdomen and concluding with what I feel is perhaps most important in 3D in the abdomen, which is workflow.

A lot of the interest in 3D ultrasound in the community of radiologists at large and especially amongst residents and fellows and to some extent amongst referring physicians, has really been spurred by CT and MRI.

I think most of you are familiar with the multiplanar reformatted and with the volumetric images that we obtained from CT and MR and certainly those are widely available now and spur people who deal with ultrasound who've only been used to looking at two dimensional images to ask themselves, why can't we do the same thing with ultrasound in the abdomen?

Now, ultrasound three dimensional ultrasound has already proven itself in a short time in areas such as OB imaging and in GYN imaging, and I will not discuss those applications, but the abdomen is still a bit of an earlier in terms of its applicability.

3D Ultrasound Instrumentation

Let's start off by talking a little bit about 3D instrumentation in ultrasound.

As you know, acquiring ultrasound for two dimensional imaging requires a transducer that images a plane of section that's illustrated here.

However, to acquire a three dimensional volume, that plane has to be moved through a three dimensional space and that's illustrated by this animation.

One of the earliest approaches to 3D ultrasound was to use something called position sensing whereby the position of the transducer was determined by three sensors which could determine its precise location in the x, y, and Z plane, and then that information would be used as the transducer was moved over the area of interest and a 3D image reconstructed.

However, that approach was really quite cumbersome and didn't gain a lot of foothold.

Over the past few years, multiple ultrasound vendors have taken a different approach to moving the 2D imaging plane through a 3D volume.

The first approach that they used, which because it was easier to accomplish, was with mechanical sweeping.

Therefore, the ultrasound transducers that were built to do mechanical sweeps had to be large enough to accommodate the physical motion of the crystals within the housing.

So the transducer that you see here you can see is physically larger than most 2D ultrasound transducers are, there are problems with mechanical sweeping, however, one of them being that anything mechanical has inertia associated with it and then inertia leads to problems.

For example, the frame rate isn't as high as one would like.

To some extent. The holy grail has been to produce transducers that do the sweeping of the ultrasound beam completely electronically, and there are a couple of vendors that have actually produced transducers like this that are now available on the market that do their movement of the ultrasound beam through the volume completely electronically.

Types of 3D Ultrasound Images

There are two types of 3D ultrasound images that are produced.

The first, which is probably more familiar to people is the traditional volumetric image that you see on the top.

This is an image of a bladder and I'll be showing more of examples of this later on in the talk, but this is a two dimensional on a two dimensional screen, a 3D representation of what the bladder would look like.

The second type of 3D image is a so-called multiplanar format that shown on the bottom, and I want to talk about that a little bit.

The multiplanar image display is I think a little bit difficult to grasp, but first, here's an example of one you'll notice it's divided into four quadrants To begin with, I'll dispense with the quadrant in the bottom right of the screen because that's where the 3D volumetric image would appear, but let's concentrate on the other three beginning with the one at the top left.

This plane is the actual acquisition plane that is, this is the plane in which the original acquisition of ultrasound was performed.

The other two planes, one of which I'm showing here, and the other of which I'm showing here are orthogonal to the acquisition plane and those are reconstructed from the originally acquired data.

Rationale for 3D Ultrasound in the Abdomen

One of the questions to answer as you approach the concept of doing 3D in the abdomen is why do it to begin with, and I think we can learn a little bit from the history of ultrasound in that regard.

This is a timeline going from 1975 to 2005 showing some of the milestones that ultrasound has reached over the years.

Now I've selected these, these are somewhat arbitrary, but they show some of the major advances in equipment that have occurred beginning with real-time ultrasound.

I think nobody would have any doubt that real-time imaging was a real advance and produced information that wasn't available with static imaging that was available before.

Then Endo cavitary imaging, which started to become used widely in the mid to late 1980s is another example that's clearly revolutionized imaging of gynecologic organs and the prostate and so on.

Color doppler introduced in the late eighties and early nineties was another example of a technology that really made a difference.

I think those of us who do ultrasound clinically now can't imagine ultrasound without color.

Doppler well, 3D introduced over the past few years is another potential such advance, but the question is still open as to how much utility will have in the abdomen.

The first question you have to ask yourself in deciding whether a new advance is of clinical use is, does it let you see things in a different way than before?

Can you get information that you just couldn't get with the technology that you had before?

Then there are several types of organs and structures that particularly lend themselves to three dimensional ultrasound imaging.

That's one of the reasons that it's been such a success in OB because the fetus is surrounded by amniotic fluid.

It's an ideal environment for producing 3D images and the same thing is true in the abdomen.

Anything that contains or is surrounded by fluid because there's such good contrast resolution between the fluid and the soft tissue that you're trying to image, you can get great 3D images.

Now that fluid doesn't have to be endogenous.

This is a case of a patient with neurofibromatosis whose arm was immersed in a water bath and when I activate this animation, you'll see as I rotate it, you can see the part of the neurofibroma, the hypoechoic part that's underneath the skin and you can see the actual deformation of the skin surface the part that we actually see externally.

Now, it should also point out that the skin image that shown here is not completely faithful to the original lesion and that's because of the way this image was acquired.

It was not acquired with a dedicated 3D ultrasound transducer.

For this I used a standard linear array, which I moved manually through space and that has some artifacts associated with it.

But the point is that you can get very high resolution three dimensional images this way.

Clinical Applications of 3D Ultrasound in the Abdomen

The bladder is another organ that lends itself very well to 3D imaging because it's filled with fluid and as I show in this multiplanar image display, as I move that blue line through the original acquisition plane, the corresponding reconstructed plane in the bottom left changes accordingly and I see the bladder in a projection that I would not have been able to acquire.

Externally this is only available through 3D reconstruction.

Here's another example of the bladder.

In fact, this is the same bladder I showed a few slides ago that's been reconstructed in 3D and again because it contains urine which produces great contrast with the soft tissue of the bladder wall, we can see the contours of the bladder wall very clearly.

And to further illustrate that point, here's a case of a bladder that is filled with urine but is also surrounded by ascitic fluid.

So not only can we see the inner contour of the bladder wall, we can see the outer contour as well in this case the original data for which was loaned to me by Philips medical systems and I did the 3D reconstruction on this.

As I begin to show the 3D animation of this bladder, you can see over in the coronal there are several protuberances That are extending into the bladder lumen and these are transitional cell lesions that were confirmed cystoscopically and again, we can now see this in three dimensions, so clearly because they're surrounded by urine which provides great contrast.

Here's another example of a bladder, sort of the opposite of the one that I just showed.

As I activate this animation, you'll see a hypoechoic spot and as I rotate the bladder so we could look down into it, you can see that that actually is the mouth of a bladder diverticulum and you can actually see through the diverticulum toward its back wall something that offers a different perspective of pathology that we could certainly see in two dimensions.

But I think this shows it more clearly and this is particularly advantageous when you're trying to show these sorts of structures or pathology to other physicians who aren't used to looking at 2D ultrasound images or to use for teaching purposes.

Vessels are other ideal structures because they're filled with flowing blood which provides good contrast to both the vessel wall and any abnormal structures within the vessel lumen.

Here's an example of a multiplanar reconstruction of an aortic dissection and as I move the plane back and forth through the acquisition plane, the corresponding axial image changes and you can see a dissection flap within the aortic lumen.

You can also see that on the reconstructed so-called sagittal plane image in the bottom left.

Here's the same set of images again, but in this case in volumetric mode I've done a volume reconstruction of this segment of the aorta.

This is looking from below.

You can see the aorta and the IVC to the right of it with the spine behind and as I pick up this volume and rotate it in space so that we're looking at it from the top down, you'll see that dissection flap within the aortic lumen and you can also see it in another vascular structure above and to the left of the aorta.

That's to the patient's left, which is the superior mesenteric artery which is also dissected.

Here's another vascular structure or an artificial vascular structure.

In this case it's a TIPS shunt within the liver shown in 3D.

Now we can certainly interrogate these very well in two dimensions, both with gray scale imaging and with color and spectral doppler, but I think doing 3D reconstruction of this case illustrated to me more than anything else ever had the curved pathway that these shunts take through the liver.

And as I rotate this, you can see that the echogenic shunt really has a curved path and this shows why it's often difficult to obtain good angle correction on the shunts as you're trying to perform Doppler interrogation of them.

Here's the same shunt again shown in three dimensional ultrasound with color and as I rotate this in space, again you can see it's curved nature.

You can also see some branches that extend toward the shunt including one branch of the right portal vein that has reversed flow, which you would expect with a functioning transjugular intrahepatic portosystemic shunt.

Some elements of the biliary tract also lend themselves very well to 3D ultrasound imaging.

For example, the gallbladder.

This is a case of a normal gallbladder, but we can see the inner contour of the wall in a way we've never been able to see before.

Again, the point comes up, does this actually provide information that we couldn't get before?

I think the answer right now is no, it doesn't, but provides a fresh perspective that may in some cases increase diagnostic confidence either that something is really normal or that it's abnormal.

Here's another case of a three-dimensional view of the gallbladder and as I rotate this, you can see an area of irregularity toward the gallbladder neck that represents sludge.

And in this next case, which was provided to me by Philips as well, a live 3D image of gallbladder sludge to effective sludge in this case, which looks very much like masses within the gallbladder lumen and another image of a gallbladder showing that we can actually see fairly small structures.

This is the gallbladder lumen and that bright echogenic spot along the anterior non-dependent gallbladder wall represents a calcified polyp.

I was actually quite surprised given the resolution limitations of 3D ultrasound that I was able to see this structure which is fairly small on the 3D reconstructed image.

Another multiplanar reconstruction of a gallbladder in this case a case of gallbladder cancer with local invasion of the liver.

And as I make this live, you can see as I pan through the gallbladder and the reconstructed plane, you can see the gallbladder lumen and the hypoechoic area and the contiguous liver that represents direct spread of the gallbladder cancer.

The GI tract, which is often filled with fluid also affords itself to 3D ultrasound imaging.

Here's a case of a normal stomach and as I begin to pan through that, you can look at both the axial plane shown in the top right and the coronal plane in the bottom left that really show the multi-layer appearance of the stomach wall.

I think quite nicely And in the 3D volumetric view from the same case, we can see the gastric lumen quite well even though it's not that greatly distended and we can see the rugal folds along the greater curve of the stomach.

At the back as I rotate this around, you can see those rugal folds.

There's no abnormal pathology on this case.

Fluid filled cysts and fluid collections in the abdomen also are potential targets for 3D ultrasound imaging including this hepatic cyst which as we pan around in 3D we can see the walls of the cyst from various perspectives.

Another fairly early case from a couple years ago showing an abscess in the right hepatic lobe and this is located near the porta hepatis.

You can see a bit of the portal vein and hepatic artery adjacent to this.

This was actually very useful to illustrate to our referring clinicians where this abscess was located.

Another example of a renal cyst.

As I activate the clip and pan through, you can see the septations within the cyst quite clearly in the other two planes and panning through in a different direction.

Looking at the C plane, which is again completely reconstructed by the software, it wasn't acquired directly by the transducer.

We can see the complexity in that cyst from another perspective in the reconstructed multiplanar image, The renal collecting system when it's distended with urine also affords itself to 3D imaging.

Here as you watch the reconstructed sagittal plane in the bottom you can see a nice coronal view of the kidney, a true coronal view showing the dilated collecting system, the dilated pelvis and a bit of the tortuous dilated ureter in this patient with hydronephrosis secondary to ureteral obstruction from cervical cancer.

Workflow and Efficiency in 3D Ultrasound

Well, so much for images and I think by now you've gotten the idea that these images are certainly able to provide maybe some increased confidence in what we were already seeing with 2D, although it's open to question whether they would allow you to make a diagnosis that you absolutely couldn't make before.

But I think in some respects the more important or significant promise of 3D imaging in the abdomen is increasing efficiency and this is one that I think the ultrasound vendors are particularly interested in.

Now, why bother doing this?

Certainly traditional methods of viewing ultrasound images such as in this ultrasound PACS system that we have at UAB have served us very well for many years, whether it's using static images as in the image of the gallbladder on the left or 2D real-time clips as we see in this clip of the gallbladder and the right that served us very well.

What can 3D possibly offer us here?

Well, in order to understand that, I think we need to talk a little bit about workflow and ultrasound.

In conventional workflow, the patient is brought into the ultrasound scanning room set up on the examining table.

The machine is set up with demographics and so on, and then there's some period that's been scanning.

Finally, the scans that are obtained, whether they're ultrasound clips or static images are brought to a radiologist or as we've used at UAB a radiologist physician's assistant for review and reading.

The challenge in developing a 3D workflow is great, and one place to start is to decide what we can learn from CT.

In CT we always begin by acquiring the axial data, creating multiplanar and volumetric images and then taking those images and sending them to the radiologist for review and interpretation.

In traditional 2D ultrasound, as I just said, we acquire static images and clips and again send those for review and interpretation.

3D might begin with just acquiring volumes and then taking those volumes and slicing them in various ways to create 2D slices in standard planes.

That would be based on imaging protocols as well as creating the volumetric 3D images of the types that have been showing you One question that arises based on CT experience shown here and the question is can we read from volumes alone?

I know in some areas of 3D volumetric imaging people are doing that.

I would say in most of my CT practice, I never do virtually never do that because I always feel that looking at the axial dataset is important and I think the answer in ultrasound is even more clear cut that there is no way that we can interpret from volumetric images alone at this point.

The other question that arises is the one of having too many images and I think anybody who does, especially CT angiography has encountered cases such as this one where you have literally thousands of axial images to review and this could take a long time.

Certainly reviewing all these images can be very, very time consuming and if we produce too many planar images from ultrasound will be in the same situation.

If you simply apply a 2D workflow to 3D, here is what you get.

You begin with the same setup, you spend some time scanning.

You'll notice that the scanning time is shorter than it was with the 2D workflow because the theory is that you would take the probe acquire volumes in several areas of interest that would occur very quickly.

You then send the patient on his or her way and then separately you would take those volumes and post-process them and subsequently review and read them as we do now.

But if the radiologist or the radiologist surrogate does that, that could take a long time much longer than their spending reviewing and reading right now and that's not a good thing.

So in 3D workflow, you would introduce an intermediate step of post-processing showed that I've shown here, and that post-processing would be done somewhere and by somebody who's yet to be identified.

And those are two important points in deciding on what an appropriate 3D workflow would be.

Who's gonna do it and where are they going to do it?

There are a couple ways of approaching that.

I think one is what I call centralized post-processing where you have say four scanning rooms and in each room you do the setup and 3D volume acquisition that I just talked about.

And then you take all those 3D data sets and you send them to a central place, much like the 3D imaging lab we've set up for CT and our department where all the post-processing is done by one or more individuals who are trained to do that.

And then the post-process images are provided to radiologists for review and reading.

The other alternative, because some of the ultrasound machines actually afford post-processing on the machine itself is to do this in a distributed manner so that in each room there would be set up and scanning post-processing done presumably by the sonographer and then the post-process images would go to a central place for review and reading.

In my opinion, however, both these suggested workflows may have problems because if you add 3D post-processing to the workflow and it takes more time to do than it took with a 2D method, than you haven't really gained anything unless you can convincingly demonstrate that the added time of post-processing shows you information that you absolutely could not get any other way.

And I think we're not quite there in the abdomen yet.

So I think we need to make sure that whatever 3D workflow we use doesn't add inordinate amount of time to the total amount of workflow.

I also alluded to the problem of creating too many images.

Several of the ultrasound vendors have introduced ways of taking the ultrasound 3D volume and slicing it in multiple ways.

I already showed you multiple multiplanar displays where you can look at the original acquisition plane and other orthogonal planes, but they don't have to be orthogonal.

You can take that 3D volume and slice it any way you want to.

And not only that, you can create as many slices as you want.

So there is a potential for image static image overload.

Here's an example of a kidney with multiple slices done with slicing a 3D volume.

And again, the question is, is having 10 slices of a kidney in one particular imaging plane too much for somebody to look at?

Here's another example of a pancreas.

We are typically used to looking at maybe three or four images or clips of the pancreas.

If you instead create 10 or 50 or a hundred static images of the pancreas, I think most radiologists would say that that is too much.

The other question that arises is when you use the paradigm that takes the patient does the 3D volumes and then you send the patient away and then do the slicing of the volume post hoc, the question is, will you be able to see every abnormality that you would've seen with this traditional technique of ultrasound imaging?

And this was illustrated to me.

In this case, this happens to be my own liver and what I did was took a 3D volume.

I happened to know at that point that I had a small hemangioma in my right lobe and I wanted to see if acquiring a volume just randomly in the right lobe and then slicing up the volume.

I wanted to know if I'd be able to see that and retrospectively knowing that the hemangioma was there, I think I was able to see it.

In fact, it's in the top row of images in the middle.

But quite honestly, if I didn't know that it was there, I probably would've missed it.

And here is a multiplanar display of the same lesion.

You can actually see. I've intersected all three planes to the point of intersection in each image shown where, where the lines cross is where the hemangioma is.

Quite honestly, by just taking that 3D volume and slicing it, I don't know that I would've found it, whereas it was very readily apparent if you did a 2D clip through that area.

So that's one area of concern and I think this is one of the most significant questions to be answered about 3D particularly in the abdomen.

And the point is that most recognition of abnormalities on ultrasound occurs during actual scanning because it requires often maneuvers such as turning the patient to their side or perhaps having them take a breath in a certain way or doing other things.

He can only do with real time images.

Those sorts of maneuvers are going to be impossible or at least very hard to duplicate by acquiring 3D volumes.

So what I'm suggesting as a hybrid sort of approach is what I call 3D workflow with pre scanning where you start off by setting up the patient as usual, doing a pre-scan of an area of interest that is real-time imaging with some clips to decide if the area is normal or abnormal.

Then doing some volume and image or clip acquisition, then doing the post-processing and then the review and reading.

And this would work best on an organ by organ approach.

So when the liver, for example, you would pre-scan the liver and at that point I think a trained sonographer or a physician can decide fairly quickly if there's anything that needs further attention or that needs further maneuvers from the patient, whether it's turning them on their side or some other maneuver or having them upright.

And then if things are normal, maybe acquiring some volumes and then slicing those volumes later if things are abnormal, however, stop acquire in addition to volumes, acquire static images and clips that document the pathology as needed and then go on to the next organ and so on and so on.

Proceeding through the abdominal exam,

Challenges in 3D Ultrasound for the Abdomen

there are a couple other problems with 3D application in the abdomen as well.

The first is a technical problem, which is resolution.

Currently 3D volume data sets in the abdomen are not isotropic.

That is the resolution is not equivalent in all three planes.

In fact, the resolution in the derived planes is considerably worse than it is in the acquired planes.

Now this is something that the ultrasound vendors are working toward.

They have not been able to do that yet.

I'm told that that is still a little bit off and I'm hopeful that eventually we will be able to do that.

And when that happens, I think the utility of 3D, particularly in the abdomen but also in other areas including small parts imaging will be much greater.

A second problem is one of PACS connectivity and image review.

Currently there is no DICOM standard for 3D ultrasound, although I'm told that that is probably around six months off.

Currently, there is no way to take these 3D data sets and view them or manipulate them on a standard PACS.

And currently what you have to do with them is use proprietary hardware and software offered by the ultrasound vendors in order to interact with the 3D data set.

Now you can take the 3D data set and create static images and you can create those volume views that I showed you and send those to PACS.

But all the interaction with the 3D volume, which is really essential to getting all the information that you need out of it, that all has to happen on a dedicated workstation for right now.

And I think that will eventually have to change.

Summary

So in summary, I think 3D ultrasound in the abdomen right now is as much a promise of potential as it is a reality unlike OB and GYN where it clearly has offered some advantages that are very clear cut and available right now in the abdomen.

It still has a little ways to go, but I'm very optimistic that some of the technical hurdles that I mentioned in the previous slide, such as going toward isotropic imaging and PACS connectivity and so on will be solved so that 3D ultrasound in the abdomen will be used to greater potential in the future.

Thank you very much.