Functional Contrast Imaging - SD

Introduction

I am David Cosgrove from Hammersmith Hospital in London, England, and it's a privilege and pleasure for me to present to you today.

My talk on functional contrast ultrasound imaging.

So in order to extract functional information from contrast agents and I should say outset, this is a unique opportunity for ultrasound because this kind of functional information is absolutely predicated on having microbubbles or some other contrast agent.

But to extract this information, we can either administer the contrast as a bolus injection, which is actually the way that it's most commonly administered anyway, and generate time intensity curves analogous to the time activity curves of nuclear medicine isotope studies.

Or we can use the destruction reperfusion technique and measure the same kinds of curves from that.

And there's an additional method that's suitable for some organs and body systems, and that is to measure transit times.

And I'll describe that in just a minute.

Time-Intensity Curves

So here's the time intensity curve.

The black lines show the raw data as it's acquired, and the irregularity of them seems to relate to cardiac output.

So there's some sensitivity to cardiac output.

And so in order to work with these curves, they need to be smoothed.

And that's what the red line shows, the smoothed curve that is used to calculate the data.

Now there's some assumptions underlying all of this field, which need to be correct.

It needs to be the case that the signal intensity is proportional to the microbubble concentration.

If not, this wouldn't work.

And I'll show you that that does apply at least over quite wide ranges of microbubble concentration.

We also need linear signal processing, so it's much better to use the raw or RF data than the video data that's been compressed.

And of course, you need a calculation package, which can either be self-made on a workstation or nowadays for all high-end machines that are contrast capable.

They offer built-in packages, which is the best way and the most convenient, just to show that the signal is proportional to the concentration of the agents.

Here is an animal study done with a very good method of measuring the actual flow and with a variety of flow rates using a clamp on the renal artery.

And they found that the correlation coefficient was very high at 0.92 with a good probability value.

So that supports that, and that was using an old agent called binx.

And here's a study that we did using Sono view, and you see that the injected volume did produce a regular increase in the area under the transit time curve, normalized for the baseline, with a good r squared value of 0.97.

Destruction-Reperfusion Technique

So now the destruction reperfusion method is illustrated in this slide.

This method was first developed for the myocardium, for myocardial perfusion, but it works well elsewhere.

And the principle is that you use a series of high intensity pulses to destroy the microbubbles, and then you revert to a low mi non-destructive imaging sequence.

And the high mi destructive pulses, of course clear the field.

And then the blood flow brings the microbubbles back in again, and it forms this exponential curve, from which you can calculate various features.

Here's what actually happens. This is a transplant kidney.

We've seen the flash where the HI mi was used, and then you could see the flow coming in from the hilum and outright to the edge of the transplant kidney indicating that it's very well perfused.

And from that sequence, the system develops a set of numbers, so we have time on the x axis and the intensity along the Y axis, and the dots are the measured values, and the gray curve is the fitted curve fitted to those values.

Two important features can be extracted from this.

One is the slope of the initial upstroke of the curve.

This is usually called the beta value, and it's proportional to the blood flow as it's coming into that cleared volume.

And then the level of the maximum intensity reached is called a.

And this relates to the blood volume within the tissue that you're looking at.

And these are two important physiological functions.

And it's especially useful to multiply a by beta because that is proportional, rather indirectly, but it's proportional to the blood flow volume, not exactly in milliliters per minute per ml, but it relates to it and can be useful just to show that this can work in non-cardiac applications.

This was a study that my colleague Chris Harvey at the Hammersmith performed, and here you can see the beta values and the a and the product of the two, for different parts of transplant kidneys.

And they broadly make sense with values that are believable.

And you can see that the probability of these being meaningful is significant.

I'd like to see this used much more widely.

And transplant kidneys might be one particularly useful application, but tumors as well, probably.

Functional Imaging

Now you can take this functional data and turn the information into images, and for some purposes that's useful, especially where the structure of the tissue or organ is heterogeneous.

So essentially you collect the data over an area.

Eventually it'd be nice to move to volume data, and then you plot the features that you've calculated, and usually they're displayed as a color coded overlay image.

And I think particularly this may be useful for complicated structures.

Here's an example of this done with a transplant kidney, same one as I showed you.

And here one of the features that we mentioned, the maximum value a has been chosen.

So this is really a blood volume map of the kidney, and you can sense that there's a difference between the cortex and the me pyramids.

And then here it's the same thing, now with the a times the beta.

So this is a blood flow map of the kidney showing in fact that the blood flow is pretty uniform throughout.

True Transit Times

Now, another opportunity that contrast gives us is a particularly special one for ultrasound, because I don't think it's very easy to do this with any other technique.

And this is to generate true transit times.

And the idea here is that you put some measuring method over a pair of vessels, the artery and the vein, and the kidney is a good example because these two vessels are very close to each other, so it's relatively easy to make that happen.

And you simply measure the time between the arrival in the artery and the arrival in the vein.

And that is a measure of the true transit time.

So here's this done for a kidney using, in this case, spectral doppler.

You put a gate over both the artery and the vein, and you use the fact that the stereo system from the scanner puts the flow of one side on the left channel in one direction.

On the left channel might be the artery and the vein on the other channel.

So you can separate the two out.

And here's a pair of images before and after injection of a contrast agent.

And you see the increase in the signal.

If I had this in real time, you'd of course see that the arterial signal increased before the venous signal increased.

And here's some results.

Actually rather small number of patients.

But looking at this effect in renal transplants, the controls are well-functioning, healthy transplants.

The purple is our patients with acute tubular necrosis and the white patients who are rejecting.

And you can see differences between the arterial venous transit times in these different conditions.

Obviously much more work is needed on that, but it's a promising approach to measuring renal function in the liver.

The same sort of thing can be done.

And essentially this is a way to measure how much of the arterial supply is being shunted between the artery and the veins.

And so a shortened transit time between the hepatic artery and the hepatic vein is found in any condition where there are shunts.

But the important ones that we encounter commonly clinically are cirrhosis, where the cirrhotic process generates arteriovenous shunts and in metastatic liver disease.

And so we'll look at examples of that.

Liver Transit Times in Cirrhosis

This is the arrangement of the method that we first used an intravenous injection of the microbubbles it's given, and a duplicate is placed over the hepatic vein, the draining vein, and a stopwatch started so that you can measure how long it takes to cross between them.

And this is the method in practice, the Doppler gate over a hepatic vein here before the contrast.

And here after, and you can see the increase in the signal, which of course you can also hear on the audio output.

Now in cirrhosis, a number of things conspire together to shorten this transit time, the most important of which are intra hepatic shunts diagrammed here, going straight from the artery to the vein, the hepatic vein.

And here is a tracing showing the normal late hepatic vein transit time with an arrival of somewhere around about 45, 50 seconds, as compared to the situation in a patient with cirrhosis where the arrival is more like 12 or 15 seconds.

So a much earlier arrival time in patients with cirrhosis.

And we looked at this in a series of patients, comparing normals with patients with biopsy proven fibrosis, and then with different degrees, A, B, and C on the child scale of cirrhosis.

And you can see a monotonic trend to earlier and earlier arrivals with more severe disease.

But of course, the Ailes show that there is overlap.

And in practice, it turns out that this method is better at picking out the more severe type levels of cirrhosis from the milder degrees of fibrosis or normal livers.

But it does seem to work.

And here are sensitivity in specificity figures for cirrhosis being really very high, not quite so good for moderate disease and less good for mild disease in a group of real patients.

Now, that method is a little bit clumsy because you have to have an operator to hold the Doppler gate in the right place and the patient's breathing and so on, and somebody else to give the injection.

And so methods to simplify that have been developed.

And this is one that was written up recently by Dr. Al.

And they used a two dimensional approach.

So they used a low mi method, and here you can see that they've centered the image on the hepatic vein, but you can see the flow arriving in the hepatic artery, and then it's appeared in the vein, and they time the difference between these.

And in this chart, you see their results looking in the same way as we looked at the degree of fibrosis minimal to severe, and again, a monotonic relationship.

But again, notice that the degrees of fibrosis do overlap, so it's not so good at distinguishing in between minor degrees of difference in fibrosis.

Liver Transit Times in Metastatic Disease

Now, as I mentioned, the same thing can be applied to malignant disease, especially metastases.

Actually hepatocellular carcinoma is not very fruitful.

I'm sure it does produce early arrival times, but since HCCs almost always arise on the basis of cirrhosis, which itself produces early arrival times, it's not very helpful diagnostically.

But in metastatic disease, it's quite a great interest.

And the idea is that the metastases themselves contain arteriovenous shunts, but in addition, they seem to secrete some sort of general agent, which opens up shunts that are normally closed off in the surrounding normal liver.

So the technique is exactly the same.

And here is our two superimposed charts, first of all, from a normal control showing the same thing as we saw earlier with the arrival somewhere around about 45 seconds compared to a patient with known metastases where the arrival was somewhere around about 10 seconds.

And my colleague Chris Harvey again performed a study on these where he took a group of patients with colorectal carcinoma and preoperatively, they were staged for their liver involvement using contrast ct, the standard.

And so they were divided on that basis into patients with overt metastases, patients whose livers seemed to be clear on ct and those with lesions that were called indeterminate.

And these are usually sub centimeter lesions that CT struggles to characterize further.

And here are the arrival times in seconds.

And you can see that almost all the metastases had early arrival times, but the patients with apparently clear livers and those with indeterminate lesions were spread about.

And some of them had early arrival times.

And in this study, we took 25 seconds as being the limit of normal.

And here, I've colored in red, the patients who on their one year follow up CT scan had developed metastasis.

And you'll see that they were all in the group with the early arrival times.

So this is quite an exciting preliminary result, because it may be that these patients with the early arrival times are the ones who should be treated with adjuvant chemotherapy, and the ones with the late arrival times maybe could be spared the inconvenience and damage that the chemotherapy causes.

Built-in Systems and Modern Applications

Now, earlier on, I mentioned that modern systems have these kinds of methods built into them.

And here's an example of this realtime sequence actually of a focal nodular hyperplasia.

You see the supply artery and the way the tumor itself, this benign tumor lights up very early before the rest of the liver.

And then it provided you collect the data correctly, and that's why we've used this slightly strange color scale.

In this case, you can put regions of interest, in this case over the tumor and a reference part of the normal liver.

And then the machine will generate from that washing curve a set of values, including a measure of the goodness of fit.

So to tell you whether this is a reliable tracing or not.

And then the various indices including the a and the beta that we talked about as well as the arrival time, time to peak enhancement, and the value of the peak enhancement itself.

So this is relatively easy to do now within any of the modern scanners.

The liver transits can also be looked at in this way.

And this is a study looking, a recent study looking at patients with a gastrointestinal malignancy, some with known metastases and others without.

And the way this was done is to put regions of interest over a hepatic artery and in hepatic vein in, and this is a series of images at different times in the sequence, and then calculate the time intensity curves.

And here are the results from this study.

And these are the patients with known liver metastases and the black are the ones with no metastases.

And you can see that the patients with no meta had later arrival times, just exactly as we had shown in Dr. Harvey's study.

Conclusion

So in conclusion, microbubble functional studies clearly are feasible, and now the companies have built in the calculation packages into all of their top end machines.

You can extract indices that may have diagnostic value, and you can create images that may be useful, particularly where the pattern of the flow is very complex.

It's simple, and importantly it's radiation free.

And that's something that we should always bear in mind, that the radiation from doing this with an isotope scan or with ct are growing problems that we're increasingly aware of.

For the liver, it looks as though it may be useful for staging and detecting metastases even perhaps before they're visible on CT or ultrasound.

And it may be helpful in patients with diffuse liver disease in the kidney.

I think transplants are very exciting, but much more work needs to be done, and maybe it could be transferred also to the native kidney.

And there are many exciting opportunities also for tumors, particularly those that are being treated with anti-angiogenesis drugs.

Thank you very much.