17 MR US Fusion

Introduction to the Freehand System

I'm gonna continue on in the same vein as Daniel talking about Mr. Fusion, talking about one particular iteration of it, and that is the one generally known as av, which is the freehand system.

Advantages of the Freehand System

There are distinct advantages to performing the biopsy outside of the MRI suite. One can save a lot of time by performing the MRI at one time and performing the biopsy at a separate location and separate time. And of course, there's a lot of money that can be saved, and there's a significant urologist buy-in because the biopsy is still in their hands.

The advantages of MRI of course is that it provides the high resolution and sensitivity for cancer that we've been talking about all day. And the advantage of the ultrasound is that it provides real time guidance, and trust has been used for many years, so it's not something unfamiliar in other words.

It was a big rev. Something we realized long ago that interrupting workflow was going to be a big problem, but by not really perturbing workflow that much, we could integrate MRI into the urologist workflow without significantly disrupting them.

Challenges in Image Fusion

We of course take the MRI data, which by the way, is usually obtained in planes orthogonal to the rectum. And compare that to the transrectal ultrasound, which generally is obtained in sort of a fan beam manner.

This shows one of the problems with cognitive fusion, which is actually quite a nice name for something that's educated guesswork. These are the two modalities are really obtained in different planes, and when trying to integrate them, it really takes great experience to do that. So the electronic fusion of the images is really something that is very important.

As I showed this slide before, all of the Mr. Fusion techniques, regardless of the manufacturer, requires a step where the MRI is segmented. That segmentation is sent to an electronic workstation, an ultrasound is obtained, that is segmented, and then those two segments are fused to each other or registered to each other.

You can do that rigidly. That is, don't change the shape, just do the best guess that you can, or you can do it elastically, where you sort of try to adapt the two shapes to each other. Mathematically all techniques, all the techniques that have been described today require that registration.

Where they differ is what happens after that, because from that moment where they're registered, now you have to link the ultrasound to the MRI, which means that you need to know where the ultrasound probe is in space. And so you can do that by attaching what electromagnetic markers to it or GPS markers to the probe.

You can do that by using an articulated arm, and then you know where in space it is, or you can do that to some extent using the image itself as a means of identifying where you are. And Dr. Verma will say a little bit about that method.

Development and Electromagnetic Tracking

Very simply, and I will say that we've been working in this area since 2007. We did our first patients with a prototype in 2007, and this is the original paper.

And basically, we put a em electromagnetic marker on the ultrasound transrectal probe, and an rf or low magnetic field is generated in the biopsy suite. And there you can then identify the pitch and roll of the ultrasound as you move it in space. And if it's registered to the MRI, the MRI will move along with it.

Hardware Requirements

The hardware that you need for this, of course, is the MRI that we use A three T Phillips. And then we used a Phillips ultrasound device. We used the Aurora electromagnetic field generator and traxell was the EM guide that was put on that.

Software Integration

And for the software, what ultimately became the Euro NAV software was registered and then fused real time with the truss on the pre acquired MRI. This has now all been integrated into a package, and I don't have any financial relationship with AV at all.

In the beginning, this is what the system looked like. It was pretty klugy looking. And this was in a operating room or cysto suite that was used for the biopsies. And so we had the workstation where the fusion took place, the ultrasound device, which is conventional ultrasound, and then the EM field field generator was right here.

There were some early problems with interference from devices in the room, but we quickly found which devices created noise, and we just don't use those. They're not significant. They were monitors and things like that that just didn't need to be used or replacements could be found. So that was the system.

Workflow and Procedure

The workflow is the following in the MRI, we do our scan, we transfer the images to a workstation, very similar to what Daniel showed, and pick the targets off that. And MR is segmented. This is an automatic procedure. It's not something that you have to do.

I don't know if you noticed that in what he was showing, but there was a automatic segmentation of the prostate. It's pretty good. It misses at the top and the bottom. So there's a little bit of a correction that you can you have to do to make sure that the prostate is accurately segmented. But the automatic system does a pretty good job of identifying it.

Then that information is sent to the procedure room where the first thing that is done is a 3D truss sweep of the prostate using the devices that are illustrated. And then using the registration software, the MR is registered with a truss, and the biopsy can then proceed.

Typically, we're going to do both 12 cores and targeted of the identified lesions.

3D Ultrasound Sweep

The 3D sweep is something that just not that hard to learn. It you can't stop. It has to be a continuous sweep through, and it has to encompass all of the gland. If it doesn't encompass all the gland, you have to repeat until you get a good gland.

And then in three dimensions, it's actually quite easy or relatively easy, I should say, to segment the ultrasound, because you can see the margins are relatively sharper. So that's something that's again, automatically done. And that segmentation is then to be fused with the MR segmentation.

Registration and Fusion

Once again, this is the automatic MR segmentation with the biopsy mark in it. The 3D ultrasound volume is obtained, and then either rigid or elastic. We often try the rigid first because if it works, it doesn't take much time. It's very simple. If it doesn't make a good registration, then we'll use the elastic, which takes a few up to about a minute to register.

And then we can then proceed where the ultrasound and the MR are superimposed on each other. The target on the MR is now transposed over to the ultrasound and realtime biopsy can be performed using standard image probe the transrectal guided biopsy.

These are all the necessary steps ups. Now the registration, again, I emphasize both rigid and elastic. Everyone I noticed that if you don't have elastic, then you're not doing it right. But in fact, the rigid often is the better of the two. And so it's very important to keep an open mind with regard to how you're going to do it.

And then this is just an image of the a the needle coming in through the target. You can see the biopsy line through there, and this is the equivalent image. And so the urologist is really looking at these two images at the same time, and they're they're moving at the same time.

So you can line up on the target, on the mr, and it should line up on the ultrasound, and then the biopsy needle can be inserted.

One good advantage of doing this in an ultrasound suite as opposed to a MR suite is you don't need non-magnetic devices. When we were doing this in bore it, everything had to be specialized to be able to be used in a magnetic field. So it's a very safe and relatively quick procedure.

Motion Correction

Now, it is very important to mention that you can't just sit back after the original registration and expect everything to go well for the next 10 or 15 minutes. Patients move. And even ever so slightly, this is all done by the way, in the decubitus position. The patient is completely awake with local anesthesia given at the start of the procedure.

So in order to make sure that you're still registered, it's important to do a kind of timeout in it's several minutes in. And this is just routinized in our plan, and we reregister just to make sure that the everything is still good, and we're still on the same page.

If you don't do that, you'll get situations like this. You, the real time ultrasound, which is right here, is not registered with the MR at all. And so you really need to bring those two back together to make this an effective procedure.

So commonly we use motion correction during the these procedures, and we'll do periodic checks to see whether the registration is accurate, whether we're still aligned with the realtime ultrasound image, and if not, then reregister the image to keep it on tact.

Now, that adds only a couple minutes to the whole procedure, but it dramatically improves the accuracy of the technique.

Procedure in Practice

Here's one of my colleagues doing one of the procedures. The patient is in a lateral DCU position. They're awake, alert, talking no problem. The they do hear the click of the biopsy. That's about the only interaction between them knowing, but it gives you the opportunity to check, to make sure the anesthesia is deep enough so that they're not feeling the biopsy.

To imagine that these procedures were done without anesthesia a decade ago, is surprising, but it most patients don't have any problem with the the anesthesia, the local anesthesia. And just this just depicts some of the needles in the

now the interface that we use is very similar to what Daniel was showing, but this is the actual final interface between be that we send off to the to get fused.

So what happens is this system automatically shows you the segmentation. You can then correct it. There's a 3D map of what's what you're seeing. And if your correction creates something that doesn't look right, you can go back and correct it again.

For instance, if in your correction you overcorrect and you start having a lump coming out of the prostate, you know that that's not right, that's just not the way prostates look. So you can go back and correct again. So this is sort of a quality control piece.

And then you can identify the lesion. It's imperceptible on this image, but it shows up in the coronal and sagittal planes, and then it'll show up on this little 3D model. And that is annotated with lesion one. So lesion one will be sent as the first biopsy.

When the when the actual biopsy is obtained, the path will be labeled lesion one. And so that you can then go back and do the following. First of all, you can say lesion one had the following results on biopsy. There's no confusion there.

Number two, you can record, and I'll show you in a second. And Daniel showed exactly this, you can record where in fact the biopsy came from in the first place. So you, if you want to go back to that area in a future times, you can know exactly where where it came from and what the path was at that time.

So maintaining the whole chain of custody, if you will, of the biopsy is very important for quality improvement or just quality control. So this is a these are good features. Of course, the next lesion that you pick will be lesion two, et cetera.

Time Efficiency

How does this look in real life? Well, we do the 3D ultrasound acquisition. So this is just time in the ultrasound suite, not it doesn't count any of the mr time. The 3D ultrasound sweep is conducted, and then there's the reconstruction of the reference 3D ultrasound with the segmentation and then the semi then the registration with semi-automatic, which is either rigid or and or elastic, which takes about two minutes.

And then we start acquiring the specimens or putting seed placements as need be with occasional timeouts for this motion compensation every so often. And so sort of skin to skin, if you will, it's about 15 minutes, which isn't far off what a routine 12 core truss biopsy takes.

It is a little longer, but you gain so much from the accuracy of being able to pinpoint where you're placing the needle.

This is sort of a from our one of our earlier iterations of this showing the target on the ultrasound, the real time motion, and then the super imposition of the MRI. So in that iteration, we could switch between the ultrasound and the MRI to check.

Right now we have it on two separate screens, and you can watch the two simultaneously and then watch the as the needle is placed into the lesion, you can see we're getting off to the side. And as you get off to the side, the target gets smaller, means you're off target, you have a nice, big target, that means you're good.

And so it's a relatively easy interface to use, but the keys are good registration and checking to make sure the registration is maintained.

Accuracy of the System

One question you always get is, how accurate is this? In a phantom where there's no motion, we were able to routinely get accuracy of 2.4 millimeters.

Now in humans, it's very hard to know. I mean, you don't really you can't really put fiduciary workers take out the the specimen and and measure things. But we did a few animal studies in which we were able to verify an accuracy of about three millimeters with some motion caused by respiration.

So it's somewhere in the somewhere between three and probably four, depending on the motion that you have. There is some variability from where you think the needle is to where it actually is, but that is usually smaller than the size of the lesion that we're biopsying.

And as Daniel mentioned, one of the nice things is to eventually be able to tell you exactly where the biopsies were taken. So this is in the routine. This is a simulation, obviously, but it a routine image where you can see where the sites were.

Clinical Results and Studies

In the last few minutes, I just wanna talk about what we've done. We've now done thousands of these, over 2000 of these biopsies using the electromagnetic linker, how well does it perform compared to the blind 12 core biopsies that were that are obtained at the same time.

They're not exactly blind in the sense that everything goes through this process as well, but they're not directed by the mr. So in this in these situations, all the patients underwent this three TMRI with everything that we have been talking about. They obtained a blinded 12 core biopsy without reference to the MRI, and then targeted two cores for each lesion identified on r.

And there are a lot of studies that come have come out of this, but I just want to highlight a few because they're really very interesting.

So, when we talk about clinically significant cancer defined as greater or equal to four plus three Gleason as a function of PSA, we noticed something very interesting that for the detection of clinically significant cancer, it was really very low for PSA less than 2.5. It happened, but it was very low, and when it happened, the targeted was better. But it was a very rare event.

And even for the PSA of 2.5 to four, notice that the pickup of clinically significant tumors is not really that different, maybe even slightly better for the 12 core. But once you get into the PSA range of four to 10, there's a dramatic difference in the clinical significant detection rate versus 12 core, and that becomes astounding at greater than 10 PSA.

And here's the overall statistic. So what's really fascinating about that data is that if you choose a cutoff value of 5.4 as a PSA, you see, and this is the cumulative detection of clinically significant disease, you see that there's really not a huge difference between the blind 12 core and the targeted biopsy up to this PSA of about 5.4.

Beyond that point, the two curves separate dramatically. And this may this is sort of interesting result. It's from one of our urology fellows who who just wrote this up, I think it's now in publication. But and I think it points to the idea that maybe we can spare doing MRI in all patients, maybe for very low PSA patients.

This the the value of targeted biopsy may not be as high as we think it is, but in higher PSA patients, I think they're gonna be real dramatic results. And there'll be cost savings.

So in this study, about a third of the patients, if we'd used the the approach that we weren't gonna do MRI in low PSA patients, we would've avoided MRI in about a third of the patients. And two thirds of them would've fallen into this category.

So at the same time, looking at clinically insignificant disease for these low PSA values, the 12 core picked up more insignificant disease. And really in this sweet spot of 2.5 to four, there was a significant reduction in the amount of insignificant disease.

So that sort of counters my first argument that maybe we shouldn't be doing this in those patients. This sort of points out that although the results, you know, the sensitivities between trusts and and targeted biopsies may superficially appear the same, they're really different lesions that we're talking about.

The truss is way biased towards insignificant disease, and the targeted biopsies are biased towards clinically significant disease.

Summary

In the interest of time, I won't dwell on everything, but summarize and just say electromagnetic tracking is an accurate method of guiding needle biopsy of the prostate, and it theoretically allows a broad range of 3D ultrasound probes to be used, not just the ones we use, but pretty much every probe that has 3D capability.

I want to emphasize that keeping the images registered to each other, regardless of the technique you use, will requires attention to detail, quality of the registration, re-registration, motion correction. But altogether, this can be accomplished in a time efficient manner, and the results are a dramatically improved rate of detection of clinically significant disease in patients with elevated PSA.

Thank you.