Transcript
Thank you, Tony. I'm certainly delighted to be here. I was tempted to apologize with maybe some repetition on my talk, covering some of the ground that has been covered by the previous speakers. But, in fact, I'm delighted because maybe one of the points about Novalis here is that here we are from all around the globe, thinking about the same things, the same problems, and how to overcome them.
So my talk is about stereotactic radiosurgery for movement disorders, particularly Parkinson's and essential tremor, and really how to validate the targets. So we all sort of know what stereotactic radiosurgery is, or at least I thought I did for a while. But certainly we get rid of the vagaries of skin markings with our radiosurgery. And we use fiducials, too, to bind to a target inside the patient. Excuse me.
And we all know the process. We used non-coplanar beams, we've got machines with a high degree of mechanical stability. We use ExacTrac, something similar to almost real-time positioning updates. We use very small cones to get rid of penumbra, and we hope to do the patients some good.
And we can be very accurate about the symmetry with that, we use Mobius to check on the dose calculations and the treatment algorithms. And there's quite a comprehensive QA process involved. So, the dosimetry is well-defined. And the positioning of the target is well-defined. And there's a meningioma wrapped around the spinal cord. And we can measure the main and standard deviation, and then incorporate that into our margins.
Now, that's fine if we can see the target. And as the previous speaker mentioned, you know, we can see the target like the trigeminal nerve, and we can target it, and we can correct for distortions. But what if we can't see the target? Well, I suppose what I'm really getting at is, if we can't see the target, it's somewhat difficult to define, but it's even more difficult to verify the target. How do we prove it really is where it is? So imaging really does come into it.
Now, we started again, with a fairly basic thing. We looked at MRI distortions in the MRI units that we use to supply data for our patients. And there's basically three magnets close to my hospital, or within the hospital, and we measured the distortion in each. And we could quantify that. But what we can't quantify is when there's something in the magnet, whether it's a phantom or a patient. What happens to those distortions then?
There's no real way of measuring the distortions on a day-to-day basis. And one thing that I was surprised to learn was that the physicists or lab technicians, who look out the MRI magnets, really only check and tune them perhaps every six to eight weeks. And I have no idea whether the magnet is up to date, whether it's in tune, or whether it's not. Again, really, as a consequence of that, we use the elastic fusion for most patients. So we...even with patients in iPlan, we do the distortion correction in elements, and then reimport the data back into iPlan.
Now, my wife would argue that this is a movement disorder. But for a lot of people, movement disorders are much more serious. You know, if someone's in a nursing home, the effect of not being able to have a drink unaided is a terrible thing in terms of their quality of life. So, as in Professor DeSalvo's [SP] presentation, we talked about the anatomy for movement disorders, and the dentato-rubro tracks, etc. And we can see these sort of cartoons or these pictures. And they look very nice, and it looks very clear as to what to target. And we can use imaging sequences to maybe try and anatomically define where these targets are.
But traditionally, when Gamma Knife treatments had been used to treat movement disorders, there have been coordinate systems. And so we started by, once we had corrected the MRI scans, to look at various coordinate systems to try and give us a roadmap of where to treat. And these are how they work, you basically take coordinates from the AC-PC line and you go so many millimeters long, so many millimeters lateral, and so many millimeters above the AC-PC line.
So far, so good. But if you look at all the different coordinate systems that have been published, you can see discrepancies between them. And they're roughly in the right area. But perhaps if you're looking at a target, which is one or two millimeters away from corticospinal tract, the differences between these positions given by all these published coordinate systems do matter. So, our thought was, why not correlate the coordinate systems with tractography, which is something that Professor DeSalvo was talked about. And in fact, the publication by Gomes [SP] there was funded from his firm.
So to do that, we looked at tractography, and the Brainlab system is deterministic. In other words, it's a mathematical solution to that problem. But there are other ways of looking at that. A statistical system or probabilistic system can be used as well. But they're different, they tend to do different things. Deterministic tends to underestimate the tracts, and probabilistic probably overestimates the tracts.
And this is an example of that, and here we can see a deterministic tract. And here we can see a probabilistic tract. And they were different size, and hence the distance to our target would be quite different. You might feel that you're safe based on the distance to the corticospinal tract, the deterministic. But in the probabilistic, you could be within the target or within the organ at risk. How do you know which is real? But it really comes down to, what are these things? Are they actual anatomical entities? And they're not.
Their description's based on data, and really, what I'm getting at is how you see a tract depends on how you look for it. If you look for smooth connections between one part of the brain or another, you'll find that. But if you use different sequences, like HARDI, which accounts, in part, for branching or right angles and tracts, you see quite a different picture.
So when we look at DTI, don't think of it as an anatomical thing. It's got anatomical correlates, but really, what you're looking at is a statistical map, which links various parts of the brain to another. And it makes it somewhat difficult to use that wholly as a means of targeting for essential tremor or Parkinson's disease.
And this is another example of the problem. This is a paper, which looked at four volunteers, three different scanners, and they looked at the overlap between atlas and probabilistic-based targets. And we can see that the targets were different on which scanner was used. So it's scanner-specific, not patient-specific QA.
What do we do? Well, Professor [inaudible 00:08:33] has published on direct visualization using the wave sequence. So, we've been in communication with him. He used a 1.5 Tesla magnet. And we really have only access to 3 Tesla magnets. But nevertheless, we've been able to look at his sequence and try and optimize it for 3 Tesla, and then incorporate that into our protocol.
So, in summary, we believe that treatment outcomes, both efficacy and safety, can be enhanced by a patient's specific interpretation of anatomy. And both rely on spatially correct or corrected data. But who's responsible for that? I don't know. I don't have any control over the radiology departments I use. The data which we hope to collaborate with our targeting is really a statistical entity within us. But at least we're able to incorporate it within the stereotactic environment. But validation, I think requires independent methodology, which is relevant to the study question.
So this is essentially what we've done. We've basically discounted the rigid fusion. We've...we really find...let's give that a tick and not a cross. We really find elastic fusion very helpful. We've looked at a number of coordinate systems. I don't know which is right, probably we'll go with the method from my site, the [inaudible 00:10:11] method, just simply because we're able to compare our results with others in the literature. And I think being able to compare results is a very important concept.
So, perhaps we'll use this methodology here. The deterministic approach of Brainlab is fine, but it does tend to underestimate the tracts. The probabilistic approach, such as Mr. Tricks [SP] requires different software. And it's not enough to see a tract. You need to be able to generate a track as a DICOM object and put it into stereotactic space.
So this is hours of work, which does not happen here. So in terms of a workflow, there's a big question mark over this, I think we're struggling with that. But by using these three different methods, I think we are able to define a target and make sure it's a safe target. And then we can make the jump from just targeting in order to treat patients, hopefully, with some degree of safety and treatment efficacy.
Thank you very much.
So my talk is about stereotactic radiosurgery for movement disorders, particularly Parkinson's and essential tremor, and really how to validate the targets. So we all sort of know what stereotactic radiosurgery is, or at least I thought I did for a while. But certainly we get rid of the vagaries of skin markings with our radiosurgery. And we use fiducials, too, to bind to a target inside the patient. Excuse me.
And we all know the process. We used non-coplanar beams, we've got machines with a high degree of mechanical stability. We use ExacTrac, something similar to almost real-time positioning updates. We use very small cones to get rid of penumbra, and we hope to do the patients some good.
And we can be very accurate about the symmetry with that, we use Mobius to check on the dose calculations and the treatment algorithms. And there's quite a comprehensive QA process involved. So, the dosimetry is well-defined. And the positioning of the target is well-defined. And there's a meningioma wrapped around the spinal cord. And we can measure the main and standard deviation, and then incorporate that into our margins.
Now, that's fine if we can see the target. And as the previous speaker mentioned, you know, we can see the target like the trigeminal nerve, and we can target it, and we can correct for distortions. But what if we can't see the target? Well, I suppose what I'm really getting at is, if we can't see the target, it's somewhat difficult to define, but it's even more difficult to verify the target. How do we prove it really is where it is? So imaging really does come into it.
Now, we started again, with a fairly basic thing. We looked at MRI distortions in the MRI units that we use to supply data for our patients. And there's basically three magnets close to my hospital, or within the hospital, and we measured the distortion in each. And we could quantify that. But what we can't quantify is when there's something in the magnet, whether it's a phantom or a patient. What happens to those distortions then?
There's no real way of measuring the distortions on a day-to-day basis. And one thing that I was surprised to learn was that the physicists or lab technicians, who look out the MRI magnets, really only check and tune them perhaps every six to eight weeks. And I have no idea whether the magnet is up to date, whether it's in tune, or whether it's not. Again, really, as a consequence of that, we use the elastic fusion for most patients. So we...even with patients in iPlan, we do the distortion correction in elements, and then reimport the data back into iPlan.
Now, my wife would argue that this is a movement disorder. But for a lot of people, movement disorders are much more serious. You know, if someone's in a nursing home, the effect of not being able to have a drink unaided is a terrible thing in terms of their quality of life. So, as in Professor DeSalvo's [SP] presentation, we talked about the anatomy for movement disorders, and the dentato-rubro tracks, etc. And we can see these sort of cartoons or these pictures. And they look very nice, and it looks very clear as to what to target. And we can use imaging sequences to maybe try and anatomically define where these targets are.
But traditionally, when Gamma Knife treatments had been used to treat movement disorders, there have been coordinate systems. And so we started by, once we had corrected the MRI scans, to look at various coordinate systems to try and give us a roadmap of where to treat. And these are how they work, you basically take coordinates from the AC-PC line and you go so many millimeters long, so many millimeters lateral, and so many millimeters above the AC-PC line.
So far, so good. But if you look at all the different coordinate systems that have been published, you can see discrepancies between them. And they're roughly in the right area. But perhaps if you're looking at a target, which is one or two millimeters away from corticospinal tract, the differences between these positions given by all these published coordinate systems do matter. So, our thought was, why not correlate the coordinate systems with tractography, which is something that Professor DeSalvo was talked about. And in fact, the publication by Gomes [SP] there was funded from his firm.
So to do that, we looked at tractography, and the Brainlab system is deterministic. In other words, it's a mathematical solution to that problem. But there are other ways of looking at that. A statistical system or probabilistic system can be used as well. But they're different, they tend to do different things. Deterministic tends to underestimate the tracts, and probabilistic probably overestimates the tracts.
And this is an example of that, and here we can see a deterministic tract. And here we can see a probabilistic tract. And they were different size, and hence the distance to our target would be quite different. You might feel that you're safe based on the distance to the corticospinal tract, the deterministic. But in the probabilistic, you could be within the target or within the organ at risk. How do you know which is real? But it really comes down to, what are these things? Are they actual anatomical entities? And they're not.
Their description's based on data, and really, what I'm getting at is how you see a tract depends on how you look for it. If you look for smooth connections between one part of the brain or another, you'll find that. But if you use different sequences, like HARDI, which accounts, in part, for branching or right angles and tracts, you see quite a different picture.
So when we look at DTI, don't think of it as an anatomical thing. It's got anatomical correlates, but really, what you're looking at is a statistical map, which links various parts of the brain to another. And it makes it somewhat difficult to use that wholly as a means of targeting for essential tremor or Parkinson's disease.
And this is another example of the problem. This is a paper, which looked at four volunteers, three different scanners, and they looked at the overlap between atlas and probabilistic-based targets. And we can see that the targets were different on which scanner was used. So it's scanner-specific, not patient-specific QA.
What do we do? Well, Professor [inaudible 00:08:33] has published on direct visualization using the wave sequence. So, we've been in communication with him. He used a 1.5 Tesla magnet. And we really have only access to 3 Tesla magnets. But nevertheless, we've been able to look at his sequence and try and optimize it for 3 Tesla, and then incorporate that into our protocol.
So, in summary, we believe that treatment outcomes, both efficacy and safety, can be enhanced by a patient's specific interpretation of anatomy. And both rely on spatially correct or corrected data. But who's responsible for that? I don't know. I don't have any control over the radiology departments I use. The data which we hope to collaborate with our targeting is really a statistical entity within us. But at least we're able to incorporate it within the stereotactic environment. But validation, I think requires independent methodology, which is relevant to the study question.
So this is essentially what we've done. We've basically discounted the rigid fusion. We've...we really find...let's give that a tick and not a cross. We really find elastic fusion very helpful. We've looked at a number of coordinate systems. I don't know which is right, probably we'll go with the method from my site, the [inaudible 00:10:11] method, just simply because we're able to compare our results with others in the literature. And I think being able to compare results is a very important concept.
So, perhaps we'll use this methodology here. The deterministic approach of Brainlab is fine, but it does tend to underestimate the tracts. The probabilistic approach, such as Mr. Tricks [SP] requires different software. And it's not enough to see a tract. You need to be able to generate a track as a DICOM object and put it into stereotactic space.
So this is hours of work, which does not happen here. So in terms of a workflow, there's a big question mark over this, I think we're struggling with that. But by using these three different methods, I think we are able to define a target and make sure it's a safe target. And then we can make the jump from just targeting in order to treat patients, hopefully, with some degree of safety and treatment efficacy.
Thank you very much.