Elastography of Breast Masses - HD

Introduction

Hi, I'm Richard Barr from Youngstown, Ohio, professor of radiology at the Northeastern Ohio Medical College.

Today, I'd like to give you an overview of Breast Elastography as it is in 2012.

What is Elastography?

ELASTOGRAPHY is a new technique in ultrasound, which can provide information which was previously not available.

Elasticity imaging is based on tissue stiffness or hardness rather than on anatomy.

Elastography has the potential to quantify the dis of a lesion, which was previously judged only subjectively by physical exam.

So breast elastography is really an imaging equivalent of a physical exam.

Types of Elastography

There are two types of elastography presently available.

These are strain or displacement or compression elastography, which is based on tissue deformation from an external or patient source.

And this is qualitative.

The other technique is ShearWave imaging, which applies a push pulse that results in a sheer wave propagation that can be measured as a velocity.

And this is a quantitative method.

Physics of Strain Elastography

Let me give you just a brief physics of these two.

Very simple.

In strain elastography, if we have a jello phantom that has an almond within it, and we use a spoon to push on that, you can see that the jello changes shape, and we know from that that it's soft.

However, if you look at the almond as we press with the spoon, it doesn't change shape and therefore it's hard.

So the strain algorithm actually looks at frame to frame changes in the software.

The algorithm actually varies by manufacturers, so you need to be aware of which system you're using because the amount of compression you need to do varies depending on which manufacturer's equipment you have.

Some require very minimal changes, and actually just the patient's breathing is the actual appropriate amount to get good images where others require you to use your hand to cause some displacement to get a reasonable amount of displacement for their images.

And again, this is a qualitative and not a quantitative method, so we don't get an absolute value how stiff it is, but the images are coated either in black and white or in color based on what is the hardest to the softest.

Interpretation Challenges in Strain Elastography

So this does create a problem in interpretation because if in this case, we have a breast image where we've got a normal dense breast tissue as well as fat and muscle, fat shows up as being white or soft.

And for all my images I'm using, white is soft and black is hard.

But in this image on the right, you can see, which is almost all fatty tissue, that we do get some areas of fat that are black.

And this is because we only have fat in this image.

And the algorithm forces something to be the hardest in that image.

And in this case, this is fat.

So it's very important that you realize this and remember this when you're interpreting images.

So to avoid interpretation images, it's really helpful if you increase your field of view as large as possible, and always try to contain fat, fibro, glandular tissue, pectoralis muscle, and the lesion.

By doing this, you have fat as being the softest tissue and pectoralis muscle being the hardest tissue, if there is no cancer.

So this gives you a fat as wide fibro glandular tissue is gray and pectoralis muscle is black.

If you have a cancer, the cancer will be the hardest in the lesion.

And again, fat will show up as white.

Fibro glandular tissue will show up as gray, and the cancer will show up as black.

If you don't include pectoralis muscle when you do this, you can get a situation where the fibro glandular tissue will look very black, and you may misinterpret that as being a cancer.

As long as you have a good B mode image, one can perform strain imaging depth is really not a problem.

Lesion Characteristics in Strain Elastography

Malignant lesions in strain for breast only actually have a very unique characteristic, and that is that they appear larger on elastography while benign lesions appear smaller.

So we like to use the strain to B mode ratio that is the length of the lesion on strain divided by the length of the lesion on B mode to characterize lesions as benign or malignant.

And we can do this with very high sensitivity and specificity, and we use a ratio of one as the discriminating factor.

So anything with the ratio of one or higher, we consider to be potentially malignant, in anything with a ratio of less than one we consider to be most likely benign.

So here's some examples.

This is a six millimeter invasive ductal cancer, and you can see that it is about seven and a half millimeters on the elastography giving us a ratio of greater than one, again, predictive of a malignancy.

And in this case, we've got a approximately one centimeter fibroadenoma, which on Elastography decreases in size 2.8.

Again, a ratio of less than one therefore felt to be benign.

Tsukuba Scale

There's another scale, the wayo scale, which has been suggested using a color scheme where red is soft and blue is hard.

You can classify the lesion similar to birads on a five point scale.

So one, the lesion is soft, meaning it is mostly green or red.

The A two, the lesion has a mixed soft and hard pattern.

Three, the lesion is hard and appears smaller than on the B mode image.

And here the blue is meant to represent the image on the elastography.

And the larger green circle is the size on be mode imaging four.

The lesion is hard and appears equal in size to B mode imaging five, the lesion is hard and appears larger than it does on B mode imaging.

And then there's A BGR, which is seen in cysts where we have this tri layered pattern, which is seen in Cyt.

So in a way, this is very similar to us using the strain to be mode ratio in that we consider it a four and five as being potentially malignant and a one, two, and three as being benign.

Semi-Quantitative Strain Analysis

In a way we can do a semi-quantitative process to look at the strain data by looking at the ratio of the strain numbers for the lesion and versus fat.

This can give us a relative idea of how stiff a lesion is.

And here what we've done is we've got a lesion that's hard being blue in this color scheme, and we've got fat.

And what we can do is place a region of interest over the lesion as well as over fat and get the strain numbers for these two, and then do the ratio.

This is not approved for use in the United States at this time.

But here you can see the ratio is 8.6.

So the cancer is 8.6 stiffer than fat.

Early studies have shown that in a ratio of approximately 4.5 is the most likely cutoff between benign and malignant lesions.

Multicenter Trial on Strain Elastography

We performed a multicenter trial using strain elastography.

It was done at six international sites.

We enrolled 635 patients with an average age of 56.

We had approximately 65% benign lesions and approximately 35% malignant lesions.

And we chose all comers.

These were all patients who were scheduled for an ultrasound guided biopsy.

And as you can see, we have a normal distribution, a pathology, and our results are tabulated in this Figure.

N is the number of patients at that site.

B is the number of lesions that were benign on pathology.

B, less than one is the number of these benign lesions that had a strain to B mode length ratio of less than one.

And that gives us our specificity.

M is the number of malignant lesions on pathology.

And m greater than one is the number of the malignant lesions that had a strain to B mode ratio of equal to or greater than one.

And as you can see, if we look at our sensitivity, five of the six sites had a hundred percent sensitivity.

So that the strain imaging was very good at predicting malignancies.

One site had a 97% sensitivity.

And when we look back at this, we could see that one of these lesions was DCIS and the lesion was not really well seen on B mode imaging.

And again, if you can't see these lesions well on either B mode or on elastography to get a good ratio, we really shouldn't be using this technique.

The other another one of those lesions turned out to be two lesions on pathology.

And on B mode imaging, it was interpreted as one lesion instead of two.

So the lesion actually looked like it got smaller, because the second lesion was benign and a third lesion, the lesion actually got taller, but did not get wider and was considered to be benign.

If we look at the specificities, they ranged from approximately 75% to 95%.

And I think we did this study actually several years ago, and we've learned how to improve the technique.

And I think that those centers that were getting lower specificities were using a technique that could have been improved.

And I think now that we think that the specificity is more in the range of 85 to 90%, if we look at the strain to be mode ratios and benign lesions that range from 0.2 to 1.5 with an average of 0.76, the malignant lesions had a strain ratio of 0.9 to 3.1 with an average of 1.45.

And these levels were significantly different than each other with a two tail p test of less than 0.001.

This is the gauge plot of our data, and you can see that all of the malignancies except for few, had a ratio of greater than one.

And there was a little bit overlap with benign lesions overlapping on the malignant lesions.

But overall, there's a very good distri separation of the distribution between benign and malignant.

Bullseye Artifact

Another interesting thing that we have reported is that on some vendors strain system, there's a very unique artifact which we've named the bullseye artifact.

And this occurs in cystic lesions.

And here you can see a simple cyst.

And the bullseye artifact is composed of three components.

One is a black ring with a white inner dot, and then a distal white dot.

And the combination of these three things is actually extremely significant in determining if something is a benign cyst.

We did a study and evaluating this artifact, and the bullseye artifact had perfect sensitivity, specificity, and a positive predictive value in determining the pathology proven benign cystic lesions.

This is seen in both simple as well as complicated benign cysts.

If the artifact in our series could have been used to exclude lesions from biopsy, our biopsy rate would have changed from 47.3 to 30.4%, which is statistically significant.

Our positive biopsy rate would have increased from 27% to 36%.

And again, which was statistically significant when one of the things we were very concerned about when we did this was would we miss any cystic neoplasms?

And this is the smallest lesion that we could find.

And this was a intraductal papillo.

But within this cyst, there's a two millimeter solid papilloma within the cyst.

And you can see that on the elastography, we can clearly see that the papillo shows up as a defect within the cyst artifact.

And we know that this is a solid component within this cyst.

Case Example: Multiple Lesions

This is another interesting case that I think describes how to interpret these images and things to be aware of when you're interpreting images.

This was a patient that came in for a screening mammogram that had a suspicious lesion.

On ultrasound, we can see this relatively iso coic lesion that corresponded to the mammographic abnormality.

And what I've done on the B mode images, I kind of say that this patient had a head to her lesion, which I've circled in red, a body of the lesion, and then a small tail that I have in the green area here on the lesion.

When we look at the ELAs gram, the main body the lesion was hard and it got larger, suspicious for a malignancy.

The tail also has gotten larger, again, suspicious for malignancy.

However, the head of the lesion is now actually extremely soft and we don't really see the lesion.

And on pathology, this body and tail turned out to be an invasive ductal cancer.

And what was the head was actually a benign fibroadenoma.

So here again, we have seen in many cases now what we thought was one lesion on be mode imaging actually turns out to be two lesions.

And you have to be aware that most pathologists will not tell you what additional benign lesions are present in the specimen when they go to surgery.

So you often don't know unless you go down and actually have the pathologist re look to see this.

And again, this can be a compounding factor when you're interpreting lesions and looking at the length measurements, because if the lesions are very similar in size and you do not identify that you may get false negative results.

ARFI Strain Imaging

Let's move on to strain imaging using aphy.

So in addition to me using patient motion or hand motion to cause the displacement of the lesions, we can also use an aphy push pulse, which can be used to move the tissue.

So this is not sheer wave imaging here.

We're using the aphy pulse to just generate the movement in the tissue, and we use the strain algorithm to interpret the images.

And again, what we see is in these cases that the lesion on B mode image gets larger on the strain imaging, again, consistent with a malignancy.

And again, another case, and this was done to see if this was less user independent.

One of the problems that we find with using this technique is the RFI push pulse in large breasts may not cause enough displacement deep in the breast.

And we actually prefer to use the standard strain imaging where we use patient motion to generate the displacement.

Because, and even in a large breast or a dense breast, we can get motion throughout the entire breast to get a good ELAs agram using the RV technique.

In lesions that are deep or in very dense breasts, we may not get enough displacement to actually get good results.

Shear Wave Imaging

Moving on to ShearWave imaging.

In this technique, we use a push pulse, which is a very low frequency high energy pulse into the breast.

And what this does is causes the tissue to vibrate.

And you can think of it as the push pulse is a stone and the sheer waves or the ripples when we throw the stone into the water.

And what we use is conventional B mode imaging to measure the ShearWave speed.

And the ShearWave speed vs is proportional to the stiffness of the tissue.

So the harder the tissue, the faster the ShearWave travels through that tissue.

And in imaging, we can do measurements in one small voxel doing a point measurement, or we can use a color map over a region of interest to get a overview of the shear wave speed over a region of the breast.

Physics of Shear Wave Imaging

Again, we're gonna use our jello phantom with the almond to show you our physics.

Here what we do is we place our ROI, we apply our push pulse, which generates the ripples in the tissue, and those shear waves are going to move at a speed that's based on the stiffness of that tissue.

And again, we're going to use our detection bemo pulses to measure that velocity.

Showing you this in a little bit different way to help you understand it, this image A is our push pulse.

We apply our push pulse, and then in our region of interest, we're going to use standard ultrasound to measure the displacement of the tissue.

And here you can see what we see on the displacement of the tissue based on the shear waves.

And here you can see the one that's closer.

We've got a peak that occurs faster than the others.

And as we get farther away from the push pulse, that peak gets farther and farther away.

And then if we plotted the peak of the curve versus the time, the slope of this line actually is the ShearWave speed.

And this ShearWave speed is what we are gonna color code to overly the on the map.

Examples in Shear Wave Imaging

Here's some examples.

This is a ous cancer.

And here you can see that we have shear wave velocities within the cancer that range from approximately 80 kilopascals to 123 kilopascals.

This is a benign fibroadenoma, and you can see that it has a measurement of approximately 17 kilo scales.

One of the problems like we've talked about before is that ShearWave generation is depth dependent.

We are limited by FDA requirements of how strong that push pulse can be.

And in our experience, if a lesion is deeper than four centimeters, one may not obtain results, and you can try repositioning the patient to bring a lesion closer to the skin.

But if you can't, what you'll see is there's no color coding in this box.

And this is telling you that the system did not pick up any shear waves and that this area is not interpretable by the ShearWave elastography.

BE1 Trial

There was a huge trial looking at shear elastography the be one trial.

And what they found was that if for B three and B four A lesions, sheer wave imaging suggested the lesions were benign.

And they used a ki Pascal number of approximately 80 to distinguish between benign and malignant.

That if it was benign, we could lower our BIRAD score by one suggesting that we can make a B three lesion a B two lesion, and we can make a B four A lesion a B three lesion.

And if the a ShearWave was positive, then we would upgrade those lesions and they would be required biopsy.

And using that, the sheer wave increased the specificity of ultrasound from 61% to 78% with no change in sensitivity and increased the positive predictive value from approximately 54% to 67%.

Problems with Shear Wave Imaging

We've noticed that there are some problems with sheer wave imaging in breast.

There are often blue cancers or soft cancers that we've seen.

And in our experience, approximately 50% of invasive ductal cancers either code is soft or do not code at all.

They there's no color.

They may have a ring of high velocity surrounding the tumor.

And again, if the sheer wave is not generated in that area, it's not color coded signaling to you that you don't get any information.

And you can't make any decision if the lesion is benign or malignant.

But unfortunately, there are several cancers that show up as being blue.

This is one such example.

Again, this was a six er millimeter invasive ductal cancer, and you can see that on shear wave elastography, the lesion shows up as very soft.

It has a slight ring of increased velocity.

However, it's still very low.

And if we look at the strain image, you could see that the lesion increase from six millimeters to almost eight millimeters, suggesting that this is a malignant lesion.

And this is very concerning, because what we really would like is on the ShearWave imaging, if we get a very low number to be really confident that the lesion is benign.

We've done some work in trying to decipher why this is happening.

And this is a case of an invasive ductal cancer, and this is one that has a ring around it of high velocity.

And I'm showing you here the sheer waves that are generated as we do this test.

And you can see that if we look in the peritumoral area where we're getting these high signals, we have these wave forms that have a little bit of noise, but we can still identify the peaks reasonably well to get accurate measurements.

However, if we look at what the shear waves look at within the tumor itself, we basically see there's a lot of displacement, but it's all noise.

And the problem with us coding this as soft is that the algorithm is interpreting this noise as a slow ShearWave speed.

And this is a fixable problem in that the algorithms need to be modified so that when we see this noise pattern that they do not color code the lesion is being soft, but again, color code it black.

So we know that we really can get an accurate measurement.

And one vendor has done this.

And you can see here's another blue cancer soft cancer if you will, but they've now generated a quality map.

And what this quality map does is tell us what is the quality of the shear waves.

And we've here used a green yellow red map where green is a very good shear wave, meaning go, it's good to interpret that.

And red means the shear waves are poor and shouldn't be interpreted.

And here you can see that although we've got a low shear wave velocity in the tumor, the quality map is telling us it's red.

Do not believe that data.

Pre-Compression Issues

One of the things that we have found to be the most problematic of why people who start with the elastography are not getting good results is something called pre compression.

And Elastography obviously looks at how these forces are change when we apply them to the breast, but the elasticity properties of these tissues change when they're compressed.

So in other words, if we have something and we apply a lot of pressure, it becomes stiffer and therefore we end up getting inaccurate results.

And I kind of say this is the heavy hand.

So if you're doing your breast ultrasound and you're using a heavy hand and pushing on the breast, you're changing the elastic properties and you're making softer tissues hard.

And what we have found was that by applying a significant amount of pre compression, and women at this point say this is less compression than on a mammogram, you can actually make fat have the same elasticity properties as a breast cancer.

So you really need to be doing these exams with a minimal touch.

Here's an example of what happens.

This actually is an invasive ductal cancer.

And again, it's one of these, actually, there's a little, you can see that there's really no coating here.

So we know that this is probably inaccurate results, but as we apply compression and we're pushing on the breast, you can see that we're making the velocities much higher.

And what I like to do to decide how much pre compression I'm doing is if we look at this Cooper's ligament, you can see let's do with a lot of pre compression.

Here's our Cooper's ligament at this depth.

As I lift up the probe, we can see that this is falling deeper and deeper into the far field.

So I'm lifting up the probe and I'm releasing the pressure from the breast.

And I use this all the time when I'm doing elastography that I stop before I do my ELAs agram, I look at something, be it a rib in the far field or a Cooper's ligament, and I lift up the probe and I do this until I can get whatever I'm looking at to far fall as far in the far field as possible and still get a reasonable image.

When you do this, your beam mode image will quality will suffer, but again, you're doing this for your elastography.

And to get good elastography results, oftentimes your B mode image is gonna suffer.

And you should be doing your B mode image separately to get the high quality B mode image separately then doing when you do the ELAs.

And just this is again, just going over what I said, pick an object in the far field, be it a rib, a cooper's ligament.

Lift the probe up and that object will start to fall in the far field.

And you continue to do this until you still have contact.

And I like to grade the amount of compression based on, as I drop this here, a rib without compression, it goes to four centimeters depth.

If I apply some pre compression, it goes to three centimeters.

Then I like to say we've applied a 25% pre compression.

And again, we should be doing our elastography either strain or ShearWave with as close to zero pre compression as possible.

Again, this is the results of our study, and here you can see different tissues.

And as we apply different amounts of pre compression, everything begins to increase in velocities so that at an approximately 40% pre compression, even fat has the same velocities as cancers.

Just to show you, these are the shear waves and to show you what happens as we apply pre compression.

So here is no pre compression, and again, you can kind of think here's your slope.

The as we move away from the shear wave, the peak occurs later and we get less and less amplitude.

But as we apply pre compression, you can see that we're moving all these peaks closer and closer in time.

So we're getting faster and faster Sheer wave velocities.

This is to show you what happens with strain imaging.

And what we have found, if you apply minimal pre compression, and here I'm using this rib here where we have at about two centimeters, we really get a very good elastography of this epidermoid cyst.

Here, if we're at about one point seven centimeters in depth, and we're applying mild pre compression, we don't get as quite a good of ELA gram.

And what we notice is on the frame to frame differences, we get some lesions that are good images and then some that are bad.

And if we apply significant pre compression here, we're at about 1.3 centimeters in depth that all we get is noise, and we really cannot get a good ELAs gram.

And again, the same thing happens with ShearWave.

Here we're looking at a simple cyst done correctly with very minimal pre compression.

And as add, we as we add more and more compression, and again, here you can see by this rib getting closer and closer to the skin, we can see that we've made this cyst now look like it's a cancer.

So again, you should be doing all your elastography with very little pre compression.

Summary

So to summarize what we see the differences between strain and ShearWave imaging, strain imaging, we don't really have a problem with depth.

So as long as we can get a good B mode image, we can get good results, which ShearWave imaging, we've got the problem that the ShearWave only penetrates so deep, and that's usually around four centimeters.

It may be less than a patient that's got a very dense breast.

So you are gonna have problems with deep lesions in the breast.

You can try moving the patient so that the lesion is closer to the skin.

But this is still usually problematic.

Pre compression is really a big problem with both strain and she wave imaging.

And no matter what elastography you're doing, you need to apply the least amount of pre compression as possible.

In our hands, if we look at sensitivity is the lesion of cancer.

We find that strain imaging is very, very good.

And it's definitely greater than 99% sensitive.

And that size change, which I must say only occurs in breast, is a very strong indicator of a malignancy.

Because of presently with the algorithms that are used in ShearWave imaging, we have this problem with the softer blue cancers that you need to be aware of.

So we're a little bit less sensitive at this point in time, which ShearWave imaging, and again, hopefully that will improve as the manufacturers improve their algorithm.

If we look at specificity, is it benign?

We often have a problem in strain imaging because most benign lesions have very similar elastic properties to normal dense breast tissue.

And oftentimes it's very difficult to get a good length measurement because the lesion actually blends in with the normal tissue.

Where on ShearWave imaging, this is kind of the best case that shear waves occur very easily in the softer tissues, and we get good measurements.

So we think the specificity of ShearWave imaging is better than that of strain.

If you want a quantitative measurement, we can do the semi quantitative ratios to get a relative ratio of a lesion to something else, such as fat in the image, but we really can get a absolute measurement.

On ShearWave imaging, we actually do get that absolute measurement.

And we have found that bullseye artifact has been a tremendous help for us in clinical practice because we have found many lesions, even those that we thought were solid, turned out to be benign complicated cysts.

And we find a large number of isod dense complicated cysts that are lesions that you may see on MR or maybe even palpable that you normally would not see on B mode imaging.

And again, we find that extremely useful on ShearWave imaging.

In very simple cysts, shear waves do not propagate and you will not get color.

But in most complicated cysts, there is ShearWave prop propagation, and you will get a low reading, but you will not really be able to tell if it's a cyst or solid.

So to conclude, both strain and ShearWave imaging provide additional information on breast lesion characterization.

We believe that elastography should be part of the routine breast exam uniformly.

Both techniques have advantages and disadvantages.

And the disadvantages of one technique are the advantages of others.

And we really like the combination of strain and ShearWave imaging.

And it's really satisfying to us clinically when we get the same results on strain and ShearWave imaging to really increase our confidence that we're getting accurate results.

Here are just some references we'll give you a little bit of time to have those.

And I do just want to end by saying at the time of this lecture, the FDA has not approved any quantification of elastography in the United States.

And this includes the strain ratio, which we've talked about in the comparison of a lesion stiffness to background tissue of fat or to display the sheer wave velocities that I've used in this talk.

Thank you.