Principles of Elastography - HD

Introduction

Hi, my name is Brian Garra.

I'm a radiologist, shared between the Department of Veterans Affairs and the US Food and Drug Administration.

And I'm here to talk to you today about elastography, the basics and principles of operation.

I have no financial relationships that I need to report.

Objectives

The objectives of this talk are to review the development of ultrasound based elastography methods
to review the principles and capabilities of both strain and shear wave elastography
to outline the strengths and weaknesses of both methods.

And finally, to summarize some quality and accuracy related issues and methods of mitigation.

History of Tissue Hardness Estimation

Elastography is an outgrowth of tissue hardness estimation.

Tissue hardness by palpation is a very ancient technique.

It was practiced by Egyptian physicians and reported by the architect, priest and physician imhotep, possibly as early as 3000 bc, but for sure, by 2,600 BC organs that have been commonly palpated include the liver, spleen, thyroid, neck, breast, almost any organ that is within reach.

The physics were discussed by Tim Hall in 2003.

Basically to be palpable, an object generally must be harder than the tissue surrounding it.

Imaging of Tissue Hardness

What about imaging of tissue hardness?

This is not a new development.

Compression has been used to estimate the hardness of a material, and Kevin Kelly, discussed this method and presented it back in the mid 1990s.

Echo palpation is a method by which you compress and watch while you're compressing a material to see how, spongy it is.

And I'll show you an example of that.

And finally, you can jostle a material and watch and see if it, if it contains fluid, you'll see the fluid flowing inside the lesion.

Here's an example of echo palpation.

I'm sliding a paperclip over a lipoma in a per person's groin.

And you can see it squishes like a sponge, and the paperclip starts and ends at the edges of the lesion.

You can feel the, lump with your fingers, through the paperclip while you're doing this.

So it's a very good way to identify exactly what is palpable or what is the patient's concern is, and then also to determine how stiff it is.

Development of Elastography

So imaging of tissue hardness using the technique of elastography grew, grew in steps.

You can see it started with mo tissue motion tracking in the 1980s, which gradually evolved and did doppler based tissue motion tracking in 1987, which ended up growing into a technique known as Sono elasticity Imaging, which was first reported in 1988 from the University of Rochester.

Static ultrasound elastography was the first widely used method for, stiffness imaging, and that was developed and presented starting in 1991 from the University of Texas.

So basically, there are two main methods of tissue elasticity estimation.

You have the quad C static methods, also known as static or strain elastography, where you gently compress the tissue and you monitor tissue displacement.

You end up calculating strain within the image, and that's what the image is made out of.

The dynamic methods, all include some sort of vibration, which is transmitted through tissue, and it's either tracked or imaged.

Sometimes, the velocity of the wave that's being traveling through the tissue is used to estimate the stiffness of tissues, and we'll talk about that some more.

Static or Quasi-Static Methods (Strain Elastography)

So the static or quasi static methods, which includes str elastography, are performed as follows.

You first apply a general compression using an ultrasound transducer.

You can also obtain compression from normal physiologic motion, such as breathing or vessel pulsations.

And finally, the third method of applying a compression to tissue is by using a method called acoustic radiation force.

The tissue displacement is tracked by a method called speckle tracking, which is usually a cross correlation technique.

This slide demonstrates the method of cross correlation and displacement tracking.

As you can see on the top bar, we have the pre compression wave form.

And on the post compression bar, the second one down, you see that a, there's a window, that gray area right in here, which is, it slid along the waveform until it finds a match of the waveform and the shift between the pre compression and the co co post compression for that same portion of the waveform is the shift or displacement in the tissue at that point or that depth in the image.

So strain ELAs, are, are actually created by taking all those displacement values and calculating the rate of change in the tissue displacement as a function of depth.

And, then you create an image of all these data points and you end up with the ELAs in a strain.

ELA gram stiffer or harder tissues are usually displayed as dark on a gray scale image, and softer tissues are displayed as bright or on a color display.

They may be displayed in any color, but typically blue or green is used for a stiffer tissue.

Examples of Strain Elastography

Here's one of the first ELAs that was created.

This is a infiltrating ductal carcinoma in a patient.

Here's the ultrasound image, and you can see there's an area of slightly increased echogenicity and a vague shadow, but you really don't actually even see the lesion very well on the ELAs.

The lesion shows up as a large irregular dark area of increased stiffness.

You can use color to display the ELAs agram as well.

Again, here's another cancer shown here with as dark, with some shadowing behind it.

On the B mode ultrasound and on the color ELAs agram, you can see it shows up as a different color from the softer tissues, which are displayed as blue here.

One of the methods to look at the differences in size of the lesion on elastography is the color overlay method.

Here you can see a gray scale image of a lesion, and the ELAs is laid over the gray scale image with the colors denoting stiff versus hard, stiff versus soft.

So here you can see this lesion is hard and that it stretches beyond the, visible margins of the lesion on the, be mode image.

Characteristics of Strain Elastography

So some of the characteristics of strain ELAs are the images of relative stiffness similar to an MRI.

So this is very good for focal disease, but if the, abnormality covers the entire image, it's gonna be very difficult to see because you only see stiffness relative to other areas in the image.

This, for example, is a diffuse breast carcinoma that fills up the entire image.

And on the ultrasound, you just see heterogene heterogeneous material and you can't tell if there's a mass or not because the mass is actually larger than the field of view on the ELA gram.

It has a certain value of stiffness, but you don't know what that value is and you have nothing to compare it to.

So you can't tell whether there's a lesion there or not.

This is oftentimes a situation.

You have an MRI where an organ is stiffer, and unless you have another organ or another part of the body that you can compare it to, you can't make a diagnosis.

Radiation Force Imaging

So you can create a strain gram by doing, what we call a radiation force imaging.

And this is, where you send a stronger ultrasound pulse to push on the tissue, and then you interrogate with, some tracking pulses to see how far the tissue is moved in response to the pushing pulse.

So in this illustration provided courtesy of Siemens, the push pulse is the amber waves.

You'll see them, we'll run it again.

Here's the push pulse, and then the green arrows show the tracking pulses that follow immediately and track how far the tissue has been pushed by the pushing pulse.

So after you create that image, you end up with a ELAs agram, typically of just a small area of the image.

And you can see it looks very similar to a, standard, strain ELAs agram where the lesion when it's hard is dark.

Okay, that takes care of focal disease, but how do we increase the sensitivity of the technique for diffuse disease?

There are a number of approaches that can be taken.

The first thing, you could do is to reconstruct the elastic modulus from the strain data.

This is what's known as the inverse problem.

And initially it was thought to be too complicated for any reasonable clinical solution because computing power within an ultrasound scanner was too limited to even allow this to be attempted.

Even now, it can be very time consuming and you need to have a lot of information about how much force was used to press on the area of interest.

And you also need to know a lot about the boundary conditions of the material.

So even now, you have to have additional information, rather beyond what you get from a normal strain image.

You have to know how much force was applied.

Elastic Modulus Reconstruction

The second method is a little bit simpler.

You simply create an image and you take the ratio of the strain in the area where you're interested, such as a lesion and compare it to the strain, an area where you think there's normal tissue.

This method conceptually is very simple, but has not achieved widespread use probably because people have been getting fairly variable results.

And also there hasn't been a, clear guidance about how you should acquire these and what you should use for the comparison region of interest.

So getting going to the elastic modulus reconstruction, the first technique, it has improved a lot over the past 10 years from very almost indecipherable phantom images to actually reasonable in vivo images.

So far it has been performed mostly in the breast and prostate, but it remains computationally very intensive, and you do have to have information about the force applied.

So here's a couple of examples, BMO images and, actual elastic modulus images, which is the stiffness of the material, often known as young's modulus.

And here you can see the B mode image and the elastic modulus showing, soft versus hard.

And here's a focal lesion, which on the beam mode is almost the same brightness as the surrounding, material in a phantom.

And, but you can see it has quite a bit, different softness, much softer than the surrounding material on the strain image.

And then on the Young's modulus image or the actual elastic modulus image, it shows up very clearly.

Strain Ratios

So let's talk about strain elastography and strain ratios.

It, as I mentioned before, it's a relative strain computation, and you get the strain from the lesion, and then you get the strain from a nearby normal tissue such as fat muscle or another organ, and you compute the strain ratio.

A variant of this is where you put a calibrated standoff pad in between the transducer and the tissue for which you know the stiffness.

And you can take the comparison of the lesion to that, standoff pad and actually get a sort of a quantitative number for stiffness.

It's less variable than the relative strain, but it loses accuracy the deeper you go into tissue because it's farther and farther away from the standoff pad.

Here's a couple of examples of strain ratio implementations.

This is on a Hitachi system, and you can see in the lesion the strain is a certain value given here as a percentage, and the strain out here is taken for comparison, and they perform a ratio, a simple ratio.

It's possible to actually create a strain ratio image such as this example, performed by Phillips.

It's not, clinically done at these at this point in time, but there is potential here.

So I mentioned that they tended to be widely variable, the strain ratio calculations, and they're widely available and easy to perform, but they haven't been giving, giving great results.

And, at the FDA, we've been working to determine some of the factors that create problems with the strain ratio and seeing, working to see if we can mitigate those.

So we've been working at the FDA with with phantoms, and here you can see the problem, when you put a lesion in a material, the, the material, this is a simulation, a computer simulation of what might happen.

The lesion is here and you can see the material.

The properties of the material strain wise are altered by the fact that there's a lesion in it.

And if you put your, if you do calculation, your calculate your strain ratio by putting, getting the strain from inside the lesion and outside, and you put it in the wrong spot, you're getting it erroneous value and you can see the problem gets worse the larger the lesion is, so that all these areas are gonna give you incorrect numbers and it actually is somewhere along the diagonal.

Here is where you're gonna get a correct number for your surrounding tissue and a correct strain ratio.

Here's the model and here's what it actually looks like in a, model that, includes ultrasound wave propagation, and here's an actual clinical experiment where we scan the phantom and actually see the same artifacts in the surrounding strain.

Quality Issues in Strain Elastography

So let's move on to ELA quality lesion.

Conspicuity can be variable in ELA grams and is not an indicator of how high quality the ELA gram is.

Visual determination of whether the ELAs gram is good or poor can be quite difficult, especially in borderline cases.

And you really don't want to be using poor quality ELAs grams 'cause they can give you very erroneous results.

So a number of manufacturers provide a quality indicator on the screen, which is very useful.

And we found another method that's very useful is to carefully monitor the gray scale image.

So here's an example of a couple different quality monitors.

This is on a semen system, and you can see it gives you a quality factor readout.

And in fact, if the quality factor calculated by the machine is less than 70, it won't actually let you look at the image, it graze it out so that you can't interpret it.

The quality factor is calculated by a proprietary method, but it usually involves the quality of that, displacement estimator, the cross correlation that I showed you earlier, plus the overall amount of strain that's present in the image.

Another company here, uses a pie shaped display of strain image quality, and you can see you want as many pie sectors lit up as you can, and the more you have lit the better quality.

This the ELAs gram is

By just looking at the B mode image.

While you're doing your compression, you can get a lot of information.

What you want for an optimal compression is about something that looks like this.

And I notice that the lesion isn't moving sideways.

The tissues are only moving up and down and squishing together.

Here on this suboptimal one, you can see that, that the lesion starts here and it starts moving sideways.

That's a problem that creates a poor quality ELA agram and most estimators will try to estimate if there's sideways motion and, not allow you to use ELAs grams that are created in this fashion.

Dynamic Elastography Methods

So moving on to the dynamic elastography methods.

As I mentioned earlier, some sort of vibration is transmitted through tissue and it's either imaged or tracked while it's transmitting through the tissue.

You can do this by applying an external or internal rapid compression, or vibration.

And again, you can make motion, you can make images of the tissue motion, the velocity to the waves traveling through the tissue or even the frequency shift.

Two main methods of dynamic tissue elasticity imaging are available.

The first one is a so-called vibration imaging, where you apply an external vibration device and you image those vibrations or track them.

The second one is known as sheer wave elastography, which is currently the dominant method for doing elasticity imaging, worldwide.

You apply an internal or external compression, and then you create shear waves that traverse the tissue and then you track the velocity of those shear waves as they traverse the tissue, which gives you an indication of tissue stiffness.

Vibration Methods

The vibration methods use an external vibrator, often between five and 100 hertz, and it can be a piston or a small loudspeaker attached to the patient's skin, or you can even use internal vibrations, from having the person hum.

This is an example of fre the fremitus method where you have the patient hum.

This is a breast cancer, the patient hums and you use doppler to evaluate the, tissue vibration as it transmits through the tissue.

And you can see that the cancer vibrates less than the surrounding tissue.

Another outgrowth of vibration imaging is sono elasticity imaging.

In this method, a modified color doppler, approach is used to track the tissue movement in response to external vibration.

And it relies on the principle that stiff materials vibrate less than soft materials.

Just as you saw in the fremitus example, I just presented.

Oftentimes you get a complex pattern of transmitted vibrations and by vibrating at two different frequencies, you can create either standing waves or what we call crawling waves, which are interference waves.

And the velocity of those is also dependent on tissue stiffness.

And you can gain estimates of how hard a tissue is by monitoring the velocity of those waves.

This is an example of sono elasticity imaging in a prostate cancer.

Here you can see the prostate cancer is relatively featureless on the B mode image, which is often the case, but on the sono elasticity image, it shows up as a dark area.

Shear Wave Elastography

Moving on to what we call ShearWave elastography.

We need to talk a little bit about how to track, vibrations.

As you know, there are two main wave types that can be transmitted through tissues.

One is the compressional wave, which is what we use to actually create the ultrasound image.

But this wave travels very fast, approximately 300 meters per second and is far too fast to track by an ultrasound method because the ultrasound is actually traveling at the same speed that you want to track.

Shear waves on the other hand, are, are transverse waves and they travel quite slowly through tissue and it is possible to track those using, standard ultrasound.

They're much weaker than the compressional wave.

They trans travel perpendicular to the compressional wave.

Here's an example of a compressional wave and then a shear wave.

As you can see, the compressional wave is a longitudinal wave where you have areas of compression and then relaxation.

Where the sheer wave is, is oscillating from side to side.

And you can see that by sending tracking pulses down, you can track the motion of the the tissue as it moves towards and away from the transducer.

If the transducer were up here.

This can also be done in magnetic resonance imaging.

So they use a special pulse sequence to monitor, moving waves of the tissue and they apply an external vibration.

And as they, as these transmit through tissue, their wavelength or their speed through the tissue can be monitored, and that gives you an indication of tissue stiffness.

Relating Shear Wave Speed to Tissue Stiffness

So how do you relate the speed of a ShearWave to the tissue stiffness?

Well, if in this equation it's a fairly simple equation for simple materials, here you have the ShearWave velocity or speed of the square of that times the density of the material times a a d factor, gives you the youngs modulus or tissue stiffness.

If you don't have the D factor, then you get the sheer modulus, which is also a good indicator of tissue stiffness.

But most people, have gotten used to the Young's modulus.

So for isotropic tissues and most tissues would fit this pattern.

The D is approximately three, such as in the liver and spleen.

Other tissues which are not isotropic such as muscle or kidney.

This factor is unknown, but it is not three, so it might be safer to use the shear modulus.

Systems and Examples

The first system that was clinically available to estimate liver stiffness or any stiffness of any material using shear waves is what we what is known as a FibroScan.

It has a vibrating piston.

And then onto that piston is built an ultrasound transducer.

The piston vibrates the tissue and produces the compressional wave, which is converted into a ShearWave.

As that ShearWave transmits through the tissue, the velocity is tracked by the ultrasound transducer mounted onto the uh, piston.

And the, velocity is again proportional with stiffness as I showed by the previous equation as and as the company shows on this, promotional material, it's proven to be very accurate in a large number of, studies with an area under the RC curve of approximately 0.94, corresponding to a sensitivity of approximately 90% and a specificity of approximately 96%.

So that's where the mic, a mechanical vibration applied to the tissue, both the MR and the fiber scan.

But you can also use what we call an acoustic radiation force impulse.

That's the same kind of push impulse that, I showed previously to create a strain, the ELAs gram.

But when you do that push, it also creates shear waves.

So you apply a single compression or a series of compressions using that acoustic pulse, and then you monitor the shear waves traveling away from the zone of compression, by using ultrasound tracking pulses.

And from that you can estimate the sheer wave speed.

The wave speed may be used directly as an indicator of tissue stiffness or may be used to calculate the sheer or young modulus as I showed by that equation.

So here's a graphical representation again provided by Siemens.

The push pulse is the amber wave transmitting down, well it transmitted for a second here, and then the tracking pulses of the green ones and the little squiggles of the shear waves emanating, you can see them emanating.

And then the tracking pulses monitor the position of tissue and therefore can track the, shear wave and estimate.

Its, So here's an example.

The region of interest is placed inside its lesion in the liver and, the ShearWave estimation has been performed.

So the track, the push pulse goes down through the center and then the tracking pulses are off to the side within the lesion.

And a velocity of 2.79 meters per second has been obtained.

It's also possible to make images of, of ShearWave speed.

And these are images, produced by a, a machine, called the, A IX Explorer.

And, in this, dark areas are soft and red areas are hard.

And you can see in this area of thyroid where there's multiple lesions, everything is soft.

Quality and Accuracy Issues in Shear Wave Elastography

So there is an issue with quality and accuracy with ShearWave elastography, just like there is with strain elastography, the same issues can apply to both.

In some cases, if you apply pre compression to a material, it makes it stiffer.

So if you, press too hard with your transducer before you measure either with strain or with ShearWave, you can end up with a stiffer estimate than you would otherwise.

So that's important to avoid.

Pre compression instrument dependent is very important for ShearWave because ShearWave speed depends on the frequency of the ShearWave, and different instruments produce shear waves at different frequencies, therefore their numbers are gonna be different.

So you shouldn't be comparing one instrument to another.

Patient or organ movement produces a lot of noise and decreased accuracy of the ShearWave estimates.

In addition, the ShearWave can only be created down to a certain limited depth, typically about six centimeters into tissue.

And therefore, if you try to, track at a deeper depth, you'll get erroneous answers or inaccurate answers due to too much noise.

The manufacturers prevent you from going too deep, but if you stay superficial, you'll be in a lot better position to get an accurate estimate.

And then of course there are a number of patient related factors.

The presence of ascites is a problem because she waves do not transfer.

Transmit through, fluids, inflammation and fatty infiltration can also affect your stiffness if you're looking for examples.

Fibrosis and inflammation can also stiffen the liver and interfere with that measurement.

Finally for focal lesions, reflections at the boundaries of focal lesions can, can confuse the algorithm that's tracking the ShearWave speed and give you a very erroneous estimate.

Over time, I expect that the algorithms will be improved to eliminate or reduce this problem.

Mitigation of Problems

So that gave some of the problems.

What are some of the ways to mitigate these problems?

Well, continuing research into, the factors affecting quality is underway.

As a operator, you can control the depth of acquisition, the angle of the transducer with respect to the skin and, organ of interest, the amount of pre compression you apply, and the position positioning and breathing, which are all factors.

There are, national and international consensus standards being developed, to standardize the approaches and to give you an approach that provides the most accurate measurements, with current equipment.

And finally, there's an initiative by the Radiologic Society of North America called the Quantitative Imaging Biomarker Alliance, who is looking into the issue of getting accurate measurements of liver fibrosis from sheer wave elastography and a standard protocol for that acquisition, plus estimates of how accurate your measurements will be if you follow that protocol, will be available in one to two years as a result of that effort.

Comparison of Methods

So let's compare the methods, here, say in 2014, for the two main types of elastography strain has been around since the 1990s and is widely available.

ShearWave elastography is much more limited.

There are several instruments available and more on the way image quality can be excellent with strain elastography and is good to excellent with ShearWave as well.

The maximum depth is good with strain because you can apply more force, to create images deeper into the body.

ShearWave you're limited by the attenuation of the, push pulse and by the weakness of the, shear waves.

So you're limited to six centimeters to eight centimeters.

Depth quantification is very limited on strain elastography, but excellent on sheer wave elastography.

Although the new techniques for strain elastography quantification are making, extreme progress.

Finally, strain imaging is great for focal lesions, but because, because it can't quantify very accurately is limited for diffuse disease, whereas sheer waves can be used for both diffuse and focal, but better for diffuse because of the problems, caused by reflections at the boundaries of focal lesions.

Other Strain-Based Approaches

There are a couple of other strain based approaches I wanted to mention that, have potential for the future.

One is poor elastography, which evaluates fluid movement in the interstitium over time using what we call poisson's ratio.

Finally, there's the axial shear strain method, which looks at how well a lesion is bound to the surrounding tissues, and creates images of that parameter.

This is the poor ELAs agram of a normal arm versus a arm with lymphedema.

As you can see, the normal, poor ELAs agram or sheer, or, poissons ratio image should have a value of 0.5.

And in normal tissue, that's what you get.

However, when there's fluid or edema in the tissue, the edema leaks out of the tissue and out of the field of view producing a poissons ratio less than 0.5, which you can see on these images, you can actually track how fast the image deviates from, normal into these lower poissons ratios.

It happens, it gets worse and worse over time, and you can actually track the fluid motion away from the area of pressure by using this technique.

The actual shear strain ELA gram, looks at the how tightly bound lesion is to its surrounding tissues.

In a mobile lesion such as this benign lesion, you can see that the lesion slips with relative to the surrounding tissues and produce a very high but narrow band of shear strain at the boundaries of a lesion where a lesion like a cancer, which is stuck to the surrounding tissues, can't slide.

So it produces a much broader, area of, sheer strain as it tries to drag the surrounding tissues with it under pressure.

Summary

So in summary, static elastography methods are now capable of producing high quality images of relative stiffness, and there is some progress being made, with quantification.

The newer methods such as ShearWave, produce quantification and, the issues they have with focal lesions are hopefully going to be resolved fairly soon.

Both methods, are very useful and the newer methods add to the diagnostic usefulness of ELA graphic imaging.

I should, mention that, being part of Health and Human Services, a mention of any commercial products, in connection with this presentation should not be construed as an actual or implied endorsement by the Department of Health and Human Services or the Department of Veterans Affairs.

Thank you.