Biomedical Engineering Center for Translational Research
2007 Conference and Expo
Friday, May 4, 2007.
University of Wisconsin-Madison, College of Engineering.
Engineering Centers Building.

Morning Agenda: Focus on Translational Research.
Prof. Walter Block, Departments of Biomedical Engineering, Medical Physics, and Radiology; Jessica Klaers, Kitty Moran, Youngkyoo Jung, Ethan Brodsky, Fred Kelcz, and Richard Kijowski.

Rapid 3D Isotropic MRI Cartilage and Breast Applications, 9:00 AM. 

00:00
Speaker: Prof. Robert Radwin, Chair of the Department of Biomedical Engineering. Prof. Block's research is in the magnetic resonance MR interventional procedures, MR angiography, and cardiac imaging, MR contrast mechanisms, image signal processing, and distributed computing.
Wally has a PhD from Stanford University in electrical engineering, and he has become a key member of our department.

Speaker: Prof. Block. And he is having some problems with the AV. Here we go. 
All right, thank you, Rob. This Coulter project was primarily for cartilage assessment, but I'm going to show you some other things that we've been doing in breast applications, and although some of us in this room may be unfortunate enough to experience something devastating like Parkinson's Disease or Alzheimer's, 
we're all going to experience in aging the problems with joint degradation, and we're all touched, in one way or another, through breast cancer. 
So, in advance, I'd like to thank all the students who worked with me on this project and also my clinical collaborators Fred Kelcz in the breast imaging and Rich Kijowski in the cartilage assessment.

So, in conventional 3D volumetric imaging in MR -- MR is the gold standard for joint imaging -- and MR has always touted this 3D capability. 
This is a movie (let me see if I can turn down the lights a little for that) [movie title: Conventional 3D Volumetric Imaging] this is a movie of our technique that's going to move through the joint, and  
I've tried to give you some landmarks here of the joint as we move along in this direction. 
And you can see this positive contrast, this myelographic effect, between the fluid in the joint and the cartilage here having intermediate intensity. 
The problem is to acquire this volume in MR imaging, we'll need on the order of Ny times Nz echoes; each of these blue dots on the side is an echo. And for a 256-by-256-by-256 volume, we're going to need 65,000 of these echoes. 
And in this ideal method of acquiring this type of contrast, called T2 weighting, we only get about five echoes per second.
So we're talking about a three-and-a-half hour exam. 
So that's really not the way imaging is done. 
What's done is usually to do rather thick two-dimensional slices.

And so, this project has been chosen to be one of three projects that are going to be shown through Wisconsin on the Research Channel, that's a nation-wide cable channel. 
And so I have some of the snippets from that that are being produced.
And so here we see ...
Video: "This is a image set from the conventional technique, where 26 slices are used to cover the breast. 
In our technique, we're using 320 slices. 
So you can see in this view -- the sideways view -- resolution is okay.
But because the slices are so thick to cover the breast -- and really, they have to be to use this technique to be able to complete the scan at a reasonable time -- here's the resolution in the other planes. 
You can see, very difficult to make any discernable information from them."

So, Mary Waitrovich over at DoIt is putting this together. 
And so the reason it's so slow is because MR imaging is kind of like a quarter-mile dragster. 
You load it up with fuel, you use all of the MR signal right away, and then you have to wait to refuel. 
So, what we're using is an idea that's more similar to the Prius. 
We're going to be very careful with the signal and try to maintain it. 
The neat thing about being able to maintain this with the steady-state imaging is we can maintain very high levels of signal at much faster rates. 

04:59
So in MR, fat and water have different resonant frequencies, and the neat thing with this steady-state imaging is we can kind of play with the spectral passbands -- kind of like in electrical engineering, making digital filters. 
The downside is that you see in this type of technique is that both fat and water are bright, and being able to differentiate between these is going to be key in joint imaging, because we're going to want to create contrast between the bone -- make the marrow of the bone dark, so we can see the cartilage stand out. 
And also in the breast, we don't want to be swamped by the fat signal -- we want to see the ductile tissue where most of the cancers are going to reside. 
The neat thing is there's a different phase of the signal between fat and water in these steady-state algorithms.
And so, if we look at this, if we take the signal here in this first experiment and adjust its face by 90 degrees. 
Then, if we add this experiment and this experiment, we create a fat passband, and we suppress water. 
We could do it another way. 
We could adjust the first experiment by 90 degrees the other way, and then we'd create a water passband. 
Now (let me turn the lights up a little here for this). So, the downside of this is to use these steady-state mechanisms, you have to be able to complete the experiments very quickly, on the order of two-and-a half milliseconds of these magnetic field strengths we have at our hospital, one-and-a-half Tesla. 
And conventional raster imaging wastes a lot of time preparing the magnetization and then returning it to the equilibrium. 
So there's only this small period here in the middle that you can do imaging. 
So the resolution, although we can get high signal, the resolution's very poor, on the order here of just two-and-a-half mm. 
We'd like to be far less than a millimeter in joint and breast imaging. 

So, we've developed this 3-D radial steady state imaging technique, where instead of rastering through a cube, we work on a radial coordinate system, and each of these different color schemes shows a different experiment, where we move out to the edge of the MR data space and then return. 
And so, if we look at it in two dimensions, the dataspace looks kind of like a ray of spokes, and in three dimensions, looks like a toy Koosh ball. 
So in the conventional technique, where we're only getting five echoes per second; with this, we can get on the order of 150 to 200 echoes per second and get resolution going down from two-and-a-half mm going down to the order of a third of a millimeter or a half a millimeter, depending on our magnetic field strength. 
And we've termed this VIPR-SSFP imaging. 

07:54
Here I've got another demonstration in the knee.
(I'll turn the lights down.)
Video: "The advantage of this type of imaging is we get very high resolution on both this cut, the transverse cut of the knee, and this cut, the sideways cut of the knee.
So, we can get different views of different types of degradation and different types of pathology. 
Normally, the slices are imaged so that they're designed only to be viewed in terms of one orientation, so this image would look good, but this image would look very blurry. 
Additionally, this type of technology allows us to get higher resolution than we could even in this through plane image here. 
So, now we're going to traverse across the knee. 
You can see through moving this fine structure we can see in the bone here of the femur some of the fine portions of fluid we're looking at. 
Then, as we move into the middle of the knee, we'll see some of the structures that are key for stabilizing the knee. 
Along here, you'll see the posterior cruciate ligament and along here -- the ligament that crosses between the tibia and the femur -- the anterior cruciate ligament, or ACL. 
If there was a break in there, and we saw fluid, we'd know there was a tear in that ligament. 
Next, you see very finely where these ligaments insert into the bones."

So where does the Coulter Foundation come into this? 
Well, people in MR research have known about radial techniques for some time, but they haven't been adopted clinically due to some of the concerns with implementing these over a wide number of scanners. 
So, what we've done over the years, we've solved a lot of problems that we would have had implementing this method across a wide installed base. 
So, here you see in this phantom exam of plastic and some water structures that these different experiments -- these different echoes of the outward and inward paths -- create very different images. 
They don't combine very well at all. 
In fact, they get worse, and this is because the trajectory of going through the MR data space is not exactly what we think it is, so here is what we'd ideally would think it's where we've placed it, but it might be slightly off. 
And so we can't count of the manufacturers to develop service routines to measure these things. 
Heating of the magnet could change that in itself. 
So, we developed per-patient calibration schemes, here showing the differences (turn that down again). 
On the left, moving from the left to the right, you can see this great improvement in the sharpness of the image and also the contrast between the bone and the cartilage by properly calibrating the scanner.

The second variable image quality problem that we had that we worked on with Coulter funding was off-axis imaging. 
For the patient to lie comfortably in the magnet, we can't have their knee in the center of the magnet. 
Likewise, the breast is not going to be in the center of the magnet. 
And this is another problem that's more problematic in non-Cartesian imaging such as these radial methods. 
So over the past year, we developed methods to measure these variable system delays on a per patient basis, and we patented them, published them and then, by being able to measure these timing offsets and correct for them in reconstruction, we move from [blurry] images that look like this, to this [slide shows images before correction and after correction]; again improvements in the contrast and in the sharpness. 
It's really unusable on the left.

The third aspect that we worked on with the Coulter funding was getting our technique to work on the new higher field magnets that are coming out. 
This image here in the upper left is where we started, when we started with the Coulter funding at 0.7-mm resolution on our one-and-a-half Tesla magnets. 
And then early in the project, we were able to improve the resolution to 0.56 mm, which if you think about 3 dimensions, the voxels in this image are really half the size of these voxels. 

And then over here we developed some changes to get it to work on our 3 Tesla magnets, and you can see, we're zooming in on the knee. (Let's see if we can bring that down to see those cartilage-bone interfaces.)
And then, finally, over here on the lower right, down to less than 0.4 of a millimeter.
So, over the year, we've dropped the voxel volume nearly by a factor of 8.

13:10
This shows some of the strength clinically what you can do with this technique. 

If you look in the conventional technique here, by the arrowhead, you see there is some change in the cartilage intensity, and because of the thick slices, it's difficult to realize what's going on. 

The middle image shows that we can create these "VIPR" slices in any orientation we'd like, so we can make them perpendicular to the femoral patella interface. 
And then, we can see on these reformatted slices, we see we're starting to find some divots in the cartilage, where early degradation is starting to occur.
In these examples of conventional techniques, you see this bright fluid here, and it's difficult to discern what's going on. 

It looks like, actually, the cartilage is gone, and there's just joint fluid going all of the way to the bone. 
But if we thin the slices, we can see there really is still a thin layer of cartilage here. 
What we'd like to do with this, we're setting up some clinical trials now at different hospitals across the country is to see if this one five-minute exam could replace a whole half-hour exam. 
So, the exams right now have all types of imaging to look for meniscal tears and ligament damage. 
And using some of the capability of computer workstations to reformat this cube of voxels, we can not only look at the cartilage, but then we can look at the ligaments in fine detail, too, and the meniscus. 

Finally, I'll just describe a little bit about breast imaging. 
The American Cancer Society suggested that women who have certain genetic predispositions to breast cancer get imaged every year with MRI. 
And likewise, women who have had cancer also to be imaged on a yearly basis. 
MR is great for breast cancer because it picks up every lesion, but unfortunately, it picks up a lot of benign lesions. 
So, the way MR exams are done is you do an injection of contrast and you see what lights up and you look at the way the contrast washes in and washes out. 
Unfortunately, there's a lot of different types of lesions, both benign and cancerous, that have similar signatures. 
So, we'd like to go back to this T2-weighted image here to look at the morphology to see if there are certain signatures that would tell us something is benign and we could then avoid an unnecessary biopsy. 
The problem is -- I showed you in that video before -- is that you only have one shot to look at the anatomy because in the other domains, the resolution is so blurry. 
So, Kitty Moran and Fred Kelcz have been working with this technique, and here I'll show one example here of a conventional technique. 
Here is a lesion -- blown up, can't really see much about it. 
But now, on the 3D radial method, we can take this lesion and blow it up in a number of different orientations. 
And here you can see in the axial plane the septation, and it's also visible in the coronal plane. 
And those septations are indicative of benign fiber adenomas.

So, I'd like to again thank again the Coulter Foundation for their support and thank Mary Waitrovich for her efforts with the Wisconsin Research Journal, and those who helped us with all the volunteer and patient scans. Thank you.

Speaker: Prof. Robert Radwin. Thank you.

Copyright 2007 The Board of Regents of the University of Wisconsin System.
Last modified 20-Jul-2007.
Created 20-Jul-2007.