Principles of Elastography and Tissue Strain Imaging - SD

Introduction

Hello, I'm Fleming Forsberg,
and I'm professor of radiology
and head of research at the Thomas Jefferson Research
and Education Institute.
And today I'll be talking about principles of elastography
and tissue strain imaging.

Today I'm gonna talk about the principles of elastography
and tissue strain imaging.
And this work was supported in part by the NIH.

Elasticity Imaging vs. Regular Ultrasound

When we start talking about elasticity in imaging,
obviously we contrast it to regular ultrasound imaging.
And really the difference here is
that in regular ultrasound imaging,
we are looking at the elastic properties.
At the molecular level,
we are essentially imaging the bulk modulars of the tissues.
Whereas in histography, we are looking at the properties
of tissue structure
and organization, specifically the strain modulars.

And what makes this so interesting,
seen from an images point of view, is
that we have approximately six orders of magnitude
of variation in the strain modulars over the range
of tissue structures
that you meet in the human body versus the bolt modulars,
where you're really only looking at changes within one
or at the most two orders of magnitude.
So in theory, elastography should have
a lot higher level of sensitivity.

Fundamental Equation of Elasticity Imaging

Now, this is a fundamental equation that governs elasticity imaging.
It's the young modulus given by the letter E here.
And it turns out that the Young's modulate is related
to stress over strain, where strain is the change in size of an object or rod
or whatever spring, whatever it is else you're looking at
relative to the original length.
So we are looking at changes in size displacement, essentially.
And then FORA is force over area. So it's the stress.
So what it tells you is that the elasticity is the strain
multiplied by the strain is equal to the stress.

Now, this is a quantitative parameter that's measured in kilo pascal,
but most elasticity imaging systems cannot capture this
directly, but really we choose to depict the strain instead.
So the relative displacement as a function of time
and baseline.
There are some equipment coming on the market now
that can do this, and we'll talk about this as we go on.
Strains does not have a direct physical equivalency,
but if you wanna contrast it to ultrasound imaging,
it's sort of the elasticity equivalent of echogenicity.
It's not very well defined, but we know what it is.

Basic Principles of Speckle Tracking in Elastography

So the basic idea here is that all imaging is based on doing ultrasound,
speckle tracking
and tissue motion estimation by finding out along an entire ALINE
that you see here, this is would be depth.
You see the entire a line here,
and you wanna track how this ALINE changes, in other words,
how the tissues move when there is some sort
of externally applied stress field.
So you would then obtain a second aline.
And you can see these little shifts here
where the ALINE has shifted
because the tissues have shifted.
Its those displacement that we try and measure
and depict as a function over time
to show strains in histography.

Early Work on Breast Tissues

Now, this is some of the early work on breast tissues
that was done in vitro by Cro Al out
of Baylor College in Texas.
And they looked at different pre compression rates
and different strain rates
because a lot of the oily work was done
with what's called static elastography,
where they used the transducer to put it on the sample
or tissue and apply a bit of force, and then more force,
and looking at the differences between two images, one
before compression than after to try and look
and detect those displacement.
But what they did was they did quantitative measurements.
And what you see here is that the different types
of breast tissue benigns, lesions, of course,
normal tissues, fat glandular
and fibrous tissue contrasted to ductal carcinoma
and situ DCI and the various infiltrating carcinomas.
And you see here that depending on what strain rates
and pre compression rates you're looking at,
there are certainly regions where these numbers in the cancer stand out quite dramatically from the normal tissues.

And that is essentially one of the reasons why women are asked to do self-examination of their breasts is because a lot of cancers feel hard.
There are hard lumps underneath your fingers.
And that's the principle of palpation, sort of,
using your fingers to feel changes in stress
underneath the skin.
And what elasticity
and elastography does is it moves
that concept onto an imaging mode.
So it's the image, it's the ultrasound
equivalent of palpation.

Tissue Compression Model

So if you look at it as a simple tissue compression model, you can see here
with the transducer before compression,
then you apply compression in this case with the transducer.
And then tissue structures within your region
of interest marked by the red box will have shifted.
You can see some of these will have moved down,
some of these will gotten flatter, some
of these will have moved outside the structure.
But what's important to bear in mind here is
that this extremely simple model also depends extensively on the boundary conditions, essentially.
You know, if you think about it, if you take this slap
of tissue and if you put it in a steel box, then no matter
how much compression you're gonna put at it,
objects aren't gonna move laterally like this
because the lateral motion will be constrained by the box.
So it's not enough to say
that tissues have a certain elasticity, they're stiff,
or they're soft, or they're not hard, whatever.
It also depends on what the surrounding tissues are.
Tissues are very soft,
but surrounding of heart tissue, you will not get to move it
around as much as if you have a heart tissue sitting
in a lot of soft tissue.
So it depends not only on the properties
of the tissue itself, but also on the structure
and organization of the tissues around them boundary conditions.

But once that established,
essentially be acquiring a pre compression RFA line,
and then a set time interval,
t later we'll acquire a second a line
after there's been some sort of induced motion, either
by the transducer or in external source,
or various other techniques, which we'll talk about.
But then you do a cross correlation
to find the best match within the original region
of interest to the secondary line.
And it's this shift here, this displacement over the time between firings that gives you the strain, which is what we depict in our elastic assist imaging.

Types of Algorithms in Elastography

So three basic types of algorithms that exist.
There is the compression using the transducer
as a compression paddle, either a quast static,
you push a bit, take an image, push, take another image,
which can be quite precise and quite quantitative.
But really for this to make an indent
as a clinical application that can be useful to physicians.
We are really looking at doing this in real time.
So real time implementations become key.

There are also techniques based on an external vibration source depicted with doppler ultrasound typically.
So you vibrate all the tissues,
and then that introduces local face changes,
which gets picked up by the topal ultrasound.
And you can then, again, since stiff tissues don't vibrate as much,
they will be depicted differently.
Nowadays that is most often referred to
as Sauna Elastography,
although vi elastography was another name
that was applied for techniques like this.

The third general class of algorithms is referred to
as acoustic radiation force imaging or afi.
And here we use a high intensity pulse,
but high intensity means still within the diagnostic range.
This is a pulse equivalent to what you used in color dober.
But this pulse is sent into the tissue,
and by doing that, it generates a sheer pressure,
which in turn will generate sheer waves
that will travel out laterally from the trans from the transmission direction.
And by tracking this sheer wave with low intensity pulses,
you can estimate what the sheer velocity is.
And that velocity in turn is directly proportional to the to the Youngs modular.
So you can, with afa, you can do quantitative measurements
of elastic modulars.

Challenges with Static Compression Systems

So this I think is a pre pretty clear idea why static compression systems did not make it into clinical practice.
It's a photograph of elastography equipment, ano 2002,
and you can see this huge contraption here, which was used
to generate very very precise control over the pressure being applied
to the to the patient.
So this word, well as a as a science setup for doing
excellent quantitative work,
but it was really not a suitable setup for clinical utility.
However, it can generate a lot of very interesting data.
You can see here on top phantom measurements in depicted in ultrasound sonograms,
and then on the bottom, the equivalent information depicted
in ELA grams.

Now, first of all, bear in mind
that every time you're looking at Anela gram,
you're really looking at the information from
two ultrasound images.
So in some sense just looking at ELA gram,
it'll have twice as much information as a sonogram will.
But see how the lesions that are here sitting here,
the echogenic nodules here,
you can see in the first image that there is a lesion up here.
You can see it very nicely in the,
but as you go further in, this is for different signal
to noise ratios.
And if you change those, you can see that we get more
and more acoustic shadowing, which ends up losing all information regarding the smaller lesion position posteriorly.
But in the ELAs grams, we continue
to pull enough information out.
'cause even though we can't directly depict it,
at least the differential between this image
and its pre-image is enough to pick up the location
of the demal of the deeper lesion of the of the tube.

Examples from In Vitro Studies

This is by now an old,
but still a famous image showing
ELAs obtained in vitro of a canine prostate.
So here black is soft tissues.
So you can see the gland is soft around the edges,
and you can see very nicely the the fiber structures
that holds the prostate gland together.
'cause they're stiffer and they show up as these white elements.
Here you can see the corresponding pathology images,
and then the ultrasound sonogram really doesn't show
any of this information.
And finally this is the equivalent MRI images
that shows very nicely the correlation between the MR
and the ELAs down here.

Animal Studies in Elastography

And our lab. We've been looking at animal starters for for quite a while trying to investigate some of the fundamental properties of elastography.
So we looked at canine prostates and livers.
We looked at rapid livers,
and we looked at swine with either thermal
or ethanol alcohol added in order to induce lesions.
And the concept here is
that we were mainly interested in cancer applications trying
to figure out if we could detect
or differentiate, characterize lesions based on their elasticity properties.

Here's an example of a rapid
with a six millimeter vx two tumor.
Sonogram is on your left
and the electrogram is on its right.
And what you really see is here is the lesion, and here the lesion.
So the gram depicts the lesion significantly more clearly
than the corresponding sonogram,
which is a very common feature.
And here has the pathological confirmation to show
that same six millimeter lesion.

Here's an example of one of our swine studies
where we've injected ethanol to gene, neutralize the tissues
and kill it.
You can see the corresponding gram here
and the pathological correlation that matches up very very nicely with this particular image.

Elastography for Monitoring Therapies

Given that it histography can pull
data out of shadows.
And given its sensitivity to d debate,
direct changes in tissues structure
and stiffness, it was early on postulated
that this would be an important tool
for monitoring therapies.
And this is an image from Rafael Gettis work
that was published in UMB in 2001.
You see the ELA here following FU lesion in a canine liver.
You can see the lesion very easily on the astro gram here
is on T two MRI.
But when you look at the actual sonogram,
you really cannot see it.
Whereas, of course, not surprisingly on pathology, you you see the image and the lesion very very clearly.

We've been looking at ethanol as a model
to induce hepatic lesions in swine.
Looked at five swine
with different ethanol injections, different depth.
These were imaged with an old HDI 1000 scanner.
And we looked at different scan techniques in order to depict it.
And what essentially we get are maps,
like the ones you'll see here, we get correlation maps,
displacement maps, and ELA gram.
What we are looking for is a region
with consistently high correlation.
Remember that these maps show then the direct depiction
of the correlation between the pre
and post compression a lines.
And you go in and find one
that has good high correlation where the lesion is.
You can see a nice displacement here
and then match up with the corresponding ela.
And again, you can see here the scale
for the correlation map goes from zero,
no correlation to one.
And we are looking for an area with high correlation
where the lesion is, which would be here.
And you can see displacement within this region,
but not so much the lesion would've moved
outside the imaging field.
And that's why you can depict the induced
lesion so clearly here.

Now, the ethanol induced lesions are an interesting model
to use because they really don't show up in any way, shape,
or form on traditional sonograms
that just an invisible lesion.
Whereas in later grams, they're very very clear.
Now, obviously, if you can't see them on ultrasound,
it can be difficult to find them apriori.
But with the pathology here, you can see
that once we take the liver out
and slice it up, that you do actually see the same
the same lesions as depicted here in the rogram.

And when we compared the data we obtained in these swine, we found that the area of the lesion is shown
by elastography correlated both with the ethanol dose
and with the lesion depth,
but quite satisfyingly, not with the acquisition technique.
And it wasn't dependent on the on the swine either.

Now, additionally, we did mechanical testing of specimens from these studies.
'cause we have access to the entire liver.
We could take the hepatic lesions out in some normal surrounding liver parenchyma.
And this was then sent down to Houston
where they had a,
or have a nano indenture machine,
which is essentially a very expensive piece of equipment
that forces a cylindrical,
indenture about two millimeter in diameter,
the tip of it into the tissue.
And at the same time, it uses a high resolution actuator
so it can measure the penetration and the force applied
and therefore the young modular can be calculated directly.

So here we see modular measurements for 2%
and 5% pre compression strain.
And really it's this blocky image that's the real data.
'cause this is 10 by 10 squares that the
the measurements were applied in.
But if you smooth them out a bit,
you can see quite clearly the lesion in center here
and there is a blood vessel that's somewhat annoying,
shouldn't be there, et cetera, et cetera.
But really had nothing to do with it.
What's important here is the fact that the lesions have
and on average a modulus
of about 26 kilo pascal versus about seven in the
surrounding normal liver.
So not only is that statistically very significant,
but there's almost an order of four difference between
the normal tissues and the stiffer lesion.

Sonoelastography and Vibration Techniques

Now in sono elastography
or white blood elastography as it was called once,
this is a technique that Kevin Park
and his group of in Rochester introduced to the world back in the late nineties.
But the concept is that when a small region
of elevated youngs modulars
or sheer modulars that's situated within an
otherwise homogeneous tissue is vibrated,
there is a decrease in the
vibration amplitude at the lesion.
So if you detect this local decrease, you should be able
to form images for oil grams,
but just based on a very different set of data.

Here's an example from a transverse view of a prostate.
You can see the gray scale lesion is up here
and the region you can see that is there are three big
red arrows that point at it.
'cause otherwise you really can't see it.
Whereas when you start vibrating the entire prostate,
you can see that this region here will linger as a
as a non filling defect.
You know in other words, the
the heart lesion here sits as a back black cloud,
whereas all the soft tissues that are being vibrated by the
external vibration source shows up in green on this power dolo image.

Here's another example of their work showing,
in this case a whole mount prostate obtained out
or radical prostatectomy.
Here's the pathology with the lesion done in this corner,
the equivalent sonogram, you can
to some degree see the lesion,
but then in vi histography this stands out very very nicely
as this black defect here in the
otherwise fairly uniformly green gland.

Radiation Force Based Shear Wave Excitation

Now this concept can take be taken one step further
into radiation force based shear wave excitation.
And what happens here is a high intensity
well-focused ultrasound beam is sent out
because this region of excitation,
depending on attenuation speed or sound, and
and the intensity of the bo
of the acoustic beam will generate a radiation force given
by this equation here.
And this will then start sheer wave
to propagate laterally out from the beam.
If you then fire subsequent low intensity pulses
that's shown up here, you end up with an overall overlap
between the two like depicted here.
And it's those pulses that will enable you
to track the velocity of the moving sheer waves and
therefore the associated youngs modulars.

Now the displacement that you are inducing
with the ultrasound beam at this stage are very very small.
This graph is from pulmonary paper
and ultrasonic imaging in 2006.
It's obtained in a phantom.
This is the excitation focal depth on the x axis versus the
displacement at the focal zone.
Four different attenuation factors.
And you can see that the less attenuation there is,
the more signal there is, the easier it is
to push things back.
But even when you have the lowest level of attenuation
and gets the most displacement,
we are still only looking at displacement on the order
of 10 microns and thereabouts.
And that's the sort of displacement
that we are trying to detect.
And map.

Here's an example of an RFI phantom image
where the conventional RFI is like this.
One of the problems in RFI is you cannot space the airlines too closely together
because with the long pulse,
you'll end up getting too much local heating.
But here is a four to one processing scheme
that was proposed very recently in this paper in 2007.
And you can see the improvements in data resolution relative
to the conventional RFI image here.

Clinical Applications of Elastography

There are many possible clinical applications.
The list can go on and on.
But as I mentioned, one of the things
that people focus a lot of their work on initially was cancer detection and characterization for breast lesions,
thyroid nodules, or maybe for prostate cancer detection.
But you could certainly also look at things like thrombus
and DVT or maybe the assessment of diffuse liver disease.
Is the liver cirrhotic or not?
An interesting concept is
to look at changes in the myocardium
'cause the heart is contracting and relaxing all the time.
It is sort of voluntarily applying pressure at all stages.
So there will be displacement there due to just the beating
of the heart, which can be investigated and depicted.
It would also be interesting to look at plaque hoping
to differentiate solid
and less dangerous plac from a more soft plaque
that potentially has the ability to rupture.

There are also people that are interested in musculoskeletal
evaluations because tendons ought to be fairly fairly stiff.
But if there are any rupture
or anything like that, they may go a lot more soft
and that may be a wave where histography
could naturally depict that.
And then as mentioned already previously,
we could use it for monitoring all therapies
that induce lesions
or change tissue parameters such as radiofrequency ablation
or high intensity focused ultrasound.

Commercial Elastography Systems

There are currently four manufacturers that have
illustry imaging systems for use for sale,
but I really shouldn't say imaging
'cause one of the four,
and this is echo cells from Paris, France,
that have the FibroScan device, which is a
non-imaging elasticity device.
And essentially you have a ultrasound transducer
with a piston like that and you put it on the skin
and it'll create indentations in the in the skin.
And this pressure wave will travel through the the ribs
and you essentially get a measure
of differences in propagation loc converted into Young's
modular eye so that you get a measure of how the
how the liver is looking and how the disease stages
'cause things that change the overall appearance
of the liver.
So just for example, cirrhosis could be depicted in
in this fashion
and at least the initial clinical trial shows
that this device should have an accuracy in excess
of 95%, which would be very impressive.

Elastography in Breast Lesions

That been some early developments looking at elastography
in starting the breast.
And by now there are much larger series.
But here's just to give a quick overview of some
of the lesion types that we've been looking at here.
And this was early work that Brian Garr from University
of Involve Vermont was involved in.
You can see the sonogram here with the lesion.
You can see the corresponding ELA here.
You can see the artifacts down here, deeper, deeper.
You go in more difficulties to measure, measure.

Likewise, here you see a small breast carcinoma
and you see the equivalent ELAs here.
One of the important things to consider is the fact
that the ELAs looks much bigger.
The lesion looks much bigger than the lesion looks in
the ultrasound sonogram.
And we think that has to do with the decal plastic
reactions with the cancer
and the surrounding tissues that sort
of infiltrating the tissues, making them stiffer
than they would otherwise be.
But we are just not able to depict things like
that in regular ultrasound imaging.

And here's a fiber adenoma.
Not only can you see that the size is fairly similar to
what you got originally,
but also that this lesion stays black.
There are some variations within it,
but it's not as black as as a cancer because it's a software
and more squishy benign fibroadenoma.

If you look at the paper
by Brian Gar Garr in radiology in 97,
this is his data from that study.
I'm looking at width differences versus signal strength
or higher numbers are softer and the DOA numbers are harder.
And they were trying to see if they could separate
fibroadenoma in yellow from carcinoma in in red.
And the hope is that if we could categorically say
that lesions have fall in this region would be benign,
then there would be a chance of reducing the number
of unnecessary biopsies
that are performed in the breast every year.

And Brian go, went on to try
and assist with ROC analysis the overall accuracy of breast
phy and found it to be on the order of 0.87,
as I remember 0.86

Real-Time Elastography Processing

For elastography to become a practical clinical tool.
However, we really need to move from quast static algorithms to real time freehand processing
and that is what's come online now more
and more with the different manufacturers.
There are several groups that've explored this concept.
And as I mentioned,
there are currently three different ultrasound manufacturers
that have devices on the market
that can do a stitch imaging.
And there is a fourth one which has this non-imaging device,
which was the French company.

Prostate Applications

Here's an example obtained with the Hitachi unit
of a normal human prostate gland.
You can see how soft the gland is along the edge of
of the gland here.
And you can see the the changes in speckle
and non-special in here under the central
core of the gland where there are regions
of soft tissue intermingled with more heart tissue.

If you compare examples of cancers detected with elastography, you can see here at least nine
and 10 prostate cancer that were also found in
this mode.
And in ultrasound imaging, you can see the lesions here
and you can see the lesion is here.
So you can see it's much stiffer than the surrounding
normal tissue.
Again, here contrast this to the previous one
where there were clear mixture of red and uh, blue
and green, but now it really only is more than anything.
It's a larger red area indicating stiffness.
You can see the relative lesion up here,
lesion scale from soft to to heart.

Here's a prostate cancer
that was only diagnosed with the lithography.
The arrow points out the small stiff lesion sitting here
where you can see the conventional gray scale.
And indeed, even the collar tabla did not give us any
indication that this patient had cancer at this location.

Musculoskeletal Applications

We've also looked at four patients with musculoskeletal disease just as a pilot study.
These people had essentially tennis elbows
and in one case they was symptomatic on one side compared
to the other side that was asymptomatic.
So we went in and used the probes without a compression
plate again, on the on the Hitachi system.
You can see when you look at the beam mode here,
if you contrast the left side,
you can see the tendon sitting up on top here.
And you can see that maybe in the right one it's a little
bit wider than the left one,
but it's really only when you go to the later grams
that you appreciate that the
asymptomatic tendon is very nice and stiff
and has a lot of blue in it indicating hardness.
Whereas the symptomatic one is a odd mixture of green
and red indicating all these soft tissues
because this is the tendon that has the problems.

There was also one case we studied
that had bilateral tennis elbows.
So both elbows had problems
but the left one was more than the right.
And if you look at them here, you can see how
the red collar indicating total softness is seen much more
dominantly on the left than it is on the right,
though still a mixture of red
and green on uh, on other side,
but we don't see that clear blue, harder depiction of
of the tendon that we would see in the normal images.

Conclusion

So with that in mind, I've tried to demonstrate
that you can get high quality Astros of many organs
and areas by now, but I've shown you liver, prostate,
and breast lesions and in vivo of course.
And it seems that very small liver lesions less than a
centimeter that you can't see
with a regular stenography may indeed be depicted
with very high contrast when you use histography.
And for the early observations of characterization
of breast lesions and liver standards,
there have been accuracies on RC analysis above 0.86.

So I don't know if I've installed you with a sense of awe,
but I can say thank you very much for taking the time
to listening to this lecture.