New Elastography Technologies: The Clinical Need - HD

Introduction

Hi, I am Richard Barr from Southwoods Imaging in Youngstown, Ohio.

Today, I'd like to give you a talk on some new technologies that are occurring in Elastography that may help us get more accurate results.

These are my disclosures.

Challenges in Strain Elastography

If we look at strain elastography, one of the key things that we need to do is consistency.

What are the requirements for training?

How can we come up with ways of getting people trained so we get very consistent results?

And are there some ways we can provide people as they're doing their exam to judge their accuracy on the examination so that they know which images are the most appropriate to use?

And in strain, histography, there's we use strain ratio a lot because we don't know the actual stiffness, what we need, a constant reference tissue.

And again, those are varied between literature, and I think we need to improve on what we use so that we're all using the same reference tissue and also where to take the measurements.

Challenges in ShearWave Elastography

If we look at ShearWave elastography, again, consistency is an issue.

What are the requirements for training?

Is there a way we can have the machine give you feedback so that you can judge your accuracy as you're doing the examination?

Not only to decide that this data set is good, but to show you as you're doing the examination that you may be doing something wrong and how to correct that.

Liver Stiffness Measurement

If we look at this from an organ perspective, looking at the liver, what do we need to improve our results with liver stiffness values?

So we need to standardization of techniques, and I think the SRU consensus panel has made an effort to try to standardize a lot of these things.

Where do we take the measurements? What cutoffs to use?

How many measurements should we do?

And what is the uniformity between vendors?

And in MRE Lasty, all of the vendors use the same technology, the same equipment, the same pulse sequences, and there's very little variability, if any, between the results, depending on which vendor you have.

Unfortunately, in our ultrasound techniques, that's not exactly the case.

And the kiba, QIBA committee from RSNA is working on trying to get vendors to move together and try to get more accurate results as well as more consistency between vendors.

And we'll talk about that as we move through the talk.

QIBA Standardization Efforts

So the Kiba actually wanted to try to standardize equipment.

There was a phase one which we used elastic phantoms.

And what we did was take multiple measurements in these phantoms at different stiffness values and different depths with multiple vendors at multiple sites.

And we found which vendors performance was at different stiffnesses and values, and they had that information so they could go back and again, try to move their equipment by evaluation of their algorithms to be more consistent.

Phase two has been completed and will be presented at RSNA 2015, where we use viscoelastic phantoms, which are more appropriate for liver.

So again, I think there are a lot of new ideas going on and how we can move.

And as we go forward in the future, I think you'll see that the measurements from various vendors will become more the same.

And so the variability between machines will decrease.

Understanding Variables

We need to also understand variables, which the KE book committee is doing right now.

We do not believe that steatosis affects the measurements as we're doing them now, if things change on how we exactly do these measurements that may change in the future, and again, we're trying to figure out how does inflammation affect these measurements?

And is there a way we can correct if the patient has inflammation superimposed on the fibrosis?

Because our liver stiffness values that we get actually determine liver stiffness, not fibrosis.

Depth and Frequency Effects

Again, the RV pulses have a bandwidth and they're attenuated differently at different depth.

So depending on the vendor's range of the bandwidth of the pulses that changes as we go through depth because the higher frequencies are gonna attenuate faster and the sheer wave speed is a function of the frequency.

So again, these bias tables are available for different vendors.

This is a very early one, and I think everyone has gotten better in the variability, but you can see that at different stiffness values in a different depth, depending on the vendors.

RV pulse bandwidth and frequency you can get different measurements at different locations.

And again, the vendors are aware of this, and I think we're moving over the next couple years.

I think you'll see that this effect is going to be minimized.

We talked about the kiba, and again all the vendors have been really good at participating in this and working towards a more standardization of measurements from the vendor perspective.

And again, improved algorithms to assess the accuracy of ShearWave speeds.

So they will reject poor quality measurements.

And right now we do have some of those on the system.

Some systems either give you 0 0 0 or XXX or into 2D shear waves.

You don't get a color coding on the map.

Those are all telling you that the system felt that the data that came back from the machine was not good enough to actually calculate an accurate ShearWave speed.

And I think all the vendors are working on ways of improving this so that they will reject data that is less accurate.

So we will be able to get more consistency and more good results.

Breast Elastography

If we look at breast, strain works very well for detection of breast cancers with reported sensitivities of up to 99% and specificities in the range of 85 to 90%.

The change in size on elastography with malignant lesions being larger and benign lesions being smaller is very helpful in characterization of breast masses and is very unique to breast.

And I don't think we really understand why.

But many, many studies have shown over thousands of patients that this works and is extremely accurate and is much more sensitive than any of our other ultrasound methods of characterizing breast lesions.

And is probably actually better than any other technique we have.

The strain ratio is less accurate, and this may be due to varying techniques.

We don't know exactly where we're taking the measurements.

And fat pre compression also affects values, and being able to standardize would be really helpful, improving the specificity.

Also, any other organ where we're doing strain ratio, these factors are involved.

With breast, there are some false positives and fat necrosis mastitis because of edema and some fibroadenomas and sclerotic lesions that are stiff.

And these are always false positives, and there's no way that we're going to be able to correct for them.

So our sensitivities, our specificities are always not going to be a hundred percent.

There are also some false negatives, which in our situation have all been breast lymphomas which are soft.

So again, we're never gonna have a sensitivity of a hundred percent, but the numbers we have are much better than any other technique.

So for ShearWave, breast cancers sometimes do not propagate the shear waves in an orderly fashion.

They're attenuated and refracted, and this can lead to either no color coding display or false low ShearWave speeds.

And again, the addition of a quality map by one vendor has alerted us that these measurements although they may have come up with a number are very inaccurate and shouldn't be used.

And I think this has in our practice significantly helped us in improving and eliminating false negatives.

And I think all the other vendors are now aware of this.

And as we progress we are going to have all the vendors have ways of alerting you when we have inaccurate shear waves and do not use those values in characterization of lesions.

Thyroid Elastography

For thyroid, we usually use strain ratio when we're doing strain and it's had variable results.

And we need to standardize the technique and select appropriate reference.

And some people have been using the strap muscles and some people have been using normal thyroid.

Shear waves give a quantitative measurement with no need for ratio.

However, we need to determine if the measurement is affected by the carotid pulsations.

And again, by just looking at what we have and most machines now, we really can't tell how these measurements are affected by the credit pulsation.

So for both the techniques strain and ShearWave, we need to improve our assessment of the image, the ELAs agram quality motion from the carotid breathing and swallowing.

MSK Elastography

For MSK, we have very high stiffness values of the tendons, and we really need to require modification of the sheer wave pulse sequences.

So in general, if you pick up your probe that you're using for breast, you may not get good results with MSK because the tendons are so stiff that the signals are attenuated, and we need to have more push pulses at an a narrower value to actually get good measurements.

And again, vendors need and are working on improving this so that we have specific programs or algorithms and settings that we can use that are tuned for the specific examination we're doing.

I find that when we do MSK, a large field of view is very helpful because I'd like to see when we're doing tendon work, the tendon over a long area, so I can see the difference between normal and abnormal and measure the length of the abnormality.

And one of the problems with muscles is there's an isotropic.

And by that I mean there's a ordering of the muscles and tendons, and we may get stiffness values differently depending on how we're aligned with that normal alignment of these tissues.

So again, we need to be aware of this and come up with protocols that we can eliminate this effect.

Standardizing Strain Elastography

So how do we standardize strain?

So pre compression changes the stiffness of tissues at various rates.

Applying pressure non uniformly will lead to inaccurate results.

And motion both in plane and outer plane will affect results.

So having the carotid not in your plane when you're doing a thyroid does not mean that you're having motion artifacts from the carotid, because it's still pulsating out of the plane and maybe affecting your measurements.

This is a results of pre compression that we did using various breath lesions.

And you can see that as we push with the probe that lesions become stiffer and they become stiffer at different rates.

So that if you're doing a strain ratio especially if you're using fat, because it will increase its stiffness values faster than cancers or other benign lesions, the ratio will change.

So it's really important that you're able to control pre compression.

And we do we have our technique that we standardize in our lab of applying minimal pre compression in the lab so that we're all doing things exactly the same, to eliminate that effect, to get more consistent results.

Color Maps and Visual Differences

I do wanna also mention that we have got different color maps and there's a visual appearance difference.

So I've got the same case, same actual, same raw data.

Here we are in black and white. Here we are with color.

And what we've done is superimpose the color on the gray scale map, and here's just the color alone.

And if you look at these, you can see that there appears to be differences between these.

And again, you have to be aware of this and again, standardize things in your lab and be aware of other people because if you look here where we have the thri tri laminate color pattern, red, green, blue, I think it's much easier picked up in this view than in this view.

So be aware of that ratios.

What tissue should we use as reference?

Again, we need to have some consensus as to how we do things in a very, very tight specific protocol.

To get uniform measurements between not only at your site, but across sites around the world.

And again, strain ratios may vary by vendor and strain ratio, shear wave speed, kilopascal stiffness ratios and shear wave meters per second.

Stiffness ratios are not the same.

And this is because vendors do things differently with the data as it comes in.

So for strain systems, we have this, the transducer sends the RF input, the strain estimate is made by the machine, and each vendor has its own way of doing that.

So there's variability there.

So it, we analyze the data, and then once the data is analyzed, the vendors use a scale map to present the information, which again, varies by vendor.

And we'll get our output that that is our color map.

So you have to be aware that the strain ratio on one system, even if everything is controlled, may be different than the strain ratio on another system because of variability of each of the vendors.

And hopefully with time, the vendors will come to a consensus so that they're all doing things exactly the same.

So that this issue disappears.

Advanced Feedback Tools for Strain Elastography

Some ways of with strain helping us to determine are we getting good results is a motion map.

So one vendor has come up with is displays the color codes, the displacement of the tissues at each level that can be used to determine where the same stress applied to standardized selection of the reference.

So here again, we can tell if someone is applying too much pre compression or if the compression is not uniform, because again, we want to take measurements when we're doing strain ratios where the stress is the same and the reference and the organ.

And I think this can be a really good as a training technique to visually demonstrate in real time to the users that they have a good technique.

So this is the motion map.

What's plotted here on the left is not our strain ELAs gram.

It's actually to displace into the tissues.

And you can see that as we move across in the same plane, we've got the same color.

We have a little bit more stiffness here because the our different color change because the stiffness of this thyroid is a little bit stiffer than the subcutaneous fat.

So we get a little bit less displacement.

But this can visually allow you to see that you're applying stress uniformly across the tissue.

And where there are differences, so this is a thyroid which is not taken near the carotid on the be mode image, but you can clearly see here's the ELAs agram, which I think if you looked at, you would not be able to tell that there are artifacts, but here on this motion map, you can clearly see the effect of the adjacent carotid artery that's changing the stiffness values.

So in this case, if we used our lesion and took a measurement here where we had motion artifacts, that's gonna vary significantly between different applications.

Or if we took the measurement here, again, we would get different values.

Some other examples here, I've healed and towed the transducer, so I'm pushing heavy on this end and not on this end.

And you can see again if you do this, that you can't take strain ratios when the color or the amount of pressure applied is different from where you're looking at your organ versus where it's at in the other tissue.

And again, I'm hoping that this will be implemented on a system in real time so that when you're doing your examination, you can actually look at this motion map and then use this as a training.

Because if you saw this pattern, you would know you're doing something wrong and heel toeing the transducer, and then you can modify that and in real time look to see that you're doing things correctly.

So again, showing you if we do these measurements at different locations here, we can see we've got different strains applied and we get a strain ratio of 7.2.

And again, different areas, we get strain ratios of 7.2 and 6.2.

And here if we do it, we can get a ratio of 1.9 versus 4.6.

So again, using this map to select 'cause here we've got these linear areas are artifacts telling you that this data, this image should not be used for calculating strain ratios.

If you apply too much pre compression, this is what happens.

So you see this red, so this is telling you that you've really applied way too much pre compression.

Now normally what's gonna happen, if you've got tissues that are varying stiffness, you may be able to get just a pattern like this instead of a uniformly light blue pattern because these tissues are stiffer and don't displace as much.

But again, if you look across the image, you can see we've got light blue here, blue green here, and red here.

So again, here, as long as we take measurements in the same depth, we'll be okay in getting a reasonable strain ratio.

Just some more examples.

So this motion map is real time.

It can be used as a training method.

And again, I'm think, I'm hoping that the vendor implies this as a real time map in addition to your real time str gram, so that you can actually get instant feedback as you're doing your examination to decide if you're applying the appropriate amount of stress, as well as applying it appropriately.

And then you can select the best frame for doing your measurements.

Other vendors are looking at techniques where you have a box, if you will, on the system.

And this box has little dotted lines.

And as you apply and release stress, it's showing you the amount that you're pushing and applying, and the optimum for their system is between these two lines.

So also when we're doing strain elastography, we can vary not only the displacement of the probe to cause the stress, but the frequency.

And what this vendor has is a box here.

And what you have is in this dimension, this is the optimal frequency.

This is the optimal displacement.

So ideally, to get the best picture, what you'd want to do is have your yellow box exactly match this purple box.

And that's telling you that you're applying the optimal strain in that image to again, give you most consistent and accurate results.

And again, you can use this as a method to detect the best image as you're moving along.

And also it's real time.

So you can look to see how your yellow box fits into the purple box and practice getting the appropriate technique for your system.

And again, here's just an example and you can see we move things differently, but you can go and find the image where you best have your yellow box filling the purple box to pick the appropriate slice to do your measurements.

Another interesting technique that a different vendor is using for breast is for calculating the strain ratio of the breast lesion to adjacent fat.

Again, there is a significant variability between users and even within a user.

So to try to eliminate that human process, what they've done is you can do your clip.

The system automatically looks at the image and selects the slice that is has the best optimal strain that you applied for that image.

You can then put an ROI within the lesion, hit go.

The system automatically finds the ROI of your lesion, and it also evaluates the adjacent fatty tissues and determines where's the best place to take fat.

So now you're getting a strain ratio that is here in this case 2.87 that has eliminated the human error from doing this.

And again, we would hope that this would help to unify or stabilize and make more consistent results within a lab or actually again, across the world.

And just another example, again, showing you that you just placed the dot and everything else is done for you.

In this case of a cancer, the liver, I'm sorry, the fat to lesion ratio is 6.56 suggestive of a malignancy.

ShearWave Imaging Techniques

Let's move on to ShearWave imaging.

So the calculation of the ShearWave speed is performed by the displacement caused by the passing ShearWave, and this is monitored by our standard B mode pulses.

So to calculate an accurate sheway speed, there needs to be adequate tissue displacement of the tissue as well as good signal to noise.

So each vendor has a rejection algorithm, and I think at this point we've got first generation rejection algorithms that evaluate the displacement and determine if the there's enough data to get an estimate of the ShearWave speed.

But I think all the vendors have realized that we can improve these algorithms to make sure that there's less noise in the algorithm and can reject more areas where the signals are inappropriate to give you a ShearWave speed.

So hopefully this will allow us to collect only good data and data points that may be inaccurate.

But the ejection algorithms are not recognizing that at this point, in the future will be rejected and again, give you more consistency in your measurements.

Propagation Maps

Another vendor and ShearWave has come up with this propagation map, which actually dis depicts the shear waves.

And ideally the ShearWave should move uniformly through.

So this case we would have t reliable results, in this case where the shear waves, the space between them is very different.

We, this is very unreliable and not to use measurements.

And within a given pattern, we can look to see here, although the signals are kind of wavy because there's different stiffnesses in this view, they're very parallel.

So these are reliable.

And here where we went to straight to crooked or not reliable.

So again, this gives you a guide to determine where you should take your measurements in the image.

So here we've got waves that are very different here and here we've got waves that are very different because we're number one two deep here, and in this case the carotid pulsation.

So we now know we should do our measurement in this area of higher reliability.

Example of a normal liver, example of color rejection around a vessel.

We've got different things here where here we've gotten more uniform measurements, but again, because these are very wavy and disturbed these measurements are not as accurate and probably a repeat exam would be appropriate.

Stiffness Ratios Across Vendors

I also wanna mention that again, all stiffness ratios are not the same.

So we have to be very careful when we're looking at the literature as to how the measurements were done.

Not only what the reference tissue was, but even which type of machine was used.

So in this breast cancer, in the same image, if we use the kilopascals and calculate a ratio between the tumor and adjacent fat, we would get a ratio of 34.6.

If we did the same image in the same location, but you'd meet it per second, it would be 6.5.

And the reason for that is the assumptions are based on the velocity squared to calculate kilopascals.

So these measurements are different depending on the speed of the ShearWave.

So we cannot relate these very easily.

And in the same case, I did the strain ratio on several machines and it ranged anywhere from five to eight.

So again, be very aware of that and you need to be very specific and vendors.

And again, hopefully as we move forward and we're aware of these things, that the vendors will be able to modify the systems so that we get more uniform results between vendors.

Future Directions: 3D Elastography

We can talk about 3D 3D strain is available and 3D shareway is available by some vendors.

Is the wobblers are being used now.

We probably would get more results if we had a solid state probe.

So hopefully those are coming.

If we had 3D probes, could we screen for breast cancer and prostate cancer?

I think the answer is yes, but the systems have to be over a wider field of view than they are now and have to be able to collect data a little bit sooner.

And as we get more and more information, do we need to have a workstation so that we can go and look at many of these other factors that they vendors are not presenting to us now to give us a better idea of the quality of the work in which data points are more accurate?

Conclusion

So to conclude, new technologies are becoming available, which may help in training and decrease the variability between measurements, improvements in rejection algorithms, and methods of determining if results are accurate are becoming available and will continue to grow so that with time, we should be getting more accurate and consistent results between labs and across the world.

Remember right now that all stiffness ratios are not the same.

And what should we use and is there a way of relating these or getting vendors to provide us information that would be the same across vendors?

I think we really need to have strict standard protocols for each organ to limit the variability of our measurements, not only within the lab, but between systems and across the world so that we all get similar results that can be compared.

Thank you very much.