Transcript
So, my name is Mike Taylor. I'm from Alliance Cancer Care, a community-based practice about an hour and a half south of here. And Brainlab has asked me to come in and tell you a little bit about our clinical experience using the Brainlab Multiple Mets app with the Elekta Agility MLC. Just a real quick outline of what I'm gonna cover. I'll talk briefly about our equipment, I'll talk a little bit about our QA, and then the strategy that we use when we do SRS, unfractionated SRS. I'll present some of our initial data and then I'll go in. Brainlab was kind enough to give me a sneak peek at version two of the software. So, we've actually gone back and recalculated some of our initial plans. Just quick disclosure, I'm here as a guest of Brainlab.
So, as you're gonna find out in a couple of slides, our program is relatively new. We've been going for about seven months. And anytime you bring up a new program, it's really important to have a good team behind you to work with. And I've been really fortunate to have that team of physicians, physicists, dosimetrists, therapists, and the rest of the team that I work with. I'd also like to give a shout-out to a couple of other people from Elekta Brainlab and SunNuclear who provided some information for this talk. So, our equipment, we have a Versa HD, and this is just a machine that we decided was gonna be mostly dedicated to stereotactic radiosurgery and SBRT. Therefore, we only have four energies. Four photon energies, we went with 6 and 10 and 6 and 10 flattening filter-free. There's a couple of philosophies behind that, but the first one for me as a busy physicist is a cut-down on commissioning time, but also cuts down on a lot of additional QA.
This particular machine comes with HexaPod 60 Couch from Elekta, obviously the Brainlab ExacTrac system, and the Agility MLC. So, the Agility MLC is kind of the heart of the delivery system. And for those that aren't familiar with this device, it's 165-millimeter leaf, and they are capable of spanning the entire 40x40 field. So, you are not limited by field size with this device. The transmission is very low at less than 0.4% and it is capable of what they call a virtual one-millimeter beamlet per aperture. So, the MLCs themselves can be positioned to better than one millimeter, and the jaws themselves can be positioned to better than one millimeter. So, we bring this program up, obviously, you go through the standard data collection process, the validation process of your model. Probably one of the more important things you can do is an end-to-end test. Brainlab does provide a really nice cranial phantom with a hidden target for you to do that. We did that test. In addition, we also were able to acquire another phantom and do a little bit of film. We looked at film obviously for its spatial resolution. Gafchromic EBT3 film, unfortunately, has a dose limitation of about 10 Gray. The good news is that I think it's Ashland Scientific now makes an extended dose range film that goes out to about 40 Gray. I think the bad news is it cost twice as much.
The software has pretty nicely configured the elements software so that you can map QA plans over to a phantom. And one of the things that we did was do point doses and we went with this SunNuclear StereoPHAN device. It's indexed to the couch. That kind of makes some of the measurements easy. And this one's milled for a small volume ion chamber. We've used a SunNuclear MapCHECK device to check fluences with the MapPHAN and had pretty good results. And then SunNuclear has just recently released their StereoPHAN, which is basically a 7.7-centimeter squared device with I believe it's 1,013 detectors. So you get a much higher detector density and increased spatial resolution. That device again fits inside their StereoPHAN.
As far as pretreatment quality assurance for single fraction stuff, we like to do a pretreatment once in a while and we look for that to be about 0.7 millimeters. We use the standard daily verification of better than one millimeter for all of our fractionated treatments. I think the real reason a lot of us are probably here are probably interested in the Multiple Mets application. I don't know if there's maybe a show in the audience how many people have seen this before, have actually done Multiple Mets, multiple isocenters. It's a very excruciating process. Time seems to slow down and things can take hours and days not only to plan this but also to deliver it depending on how much the patient can tolerate. One of the newer, I shouldn't say newer, but one of the more recent approaches is the VMAT approach. I'm sure more of us are familiar with that, single-isocenter. You optimize multiple targets at one time. I think it's a little bit quicker than the other approach. But one of the challenges you run into is something called the island blocking problem that was proposed by King.
When you have two targets that fall along the same path as the geometric delivery of the beam, the software is trying to optimize those two targets at exactly the same time. And what happens is you get an open spot in between where most likely there's either an organ at risk or a normal brain. So, you get this increase in normal brain dose. Obviously, you could work to try and geometrically avoid those approaches, but it gets more challenging, obviously, as you increase with the number of metastasis. Brainlab has taken a pretty unique approach with this, they take a dynamic conformal approach. You can treat Multiple Mets, you do it with a single-isocenter. And it's a dynamic conformal arc approach. But the way that they get around the island blocking program is that a given couch angle if there are gonna be two passes that hit the same or that hit two targets in one pass, they'll do it with two passes. So basically, the first pass, they hit the first target, the second pass, they hit the second target. And this significantly cuts down on a normal brain dose.
For us, the strategy and what we look at when we consider treating somebody with SRS or fractionated SRS, we look at the size. A lot of times post-op cavities will tend to be large, but things on the order of 3cm or larger, we look to fractionate depending on what they're next to. If it's it's next to an organ at risk, we'll fractionate. Targets especially in the Multiple Mets environment tend to have close proximity to each other causing dose bridging. And obviously, we look at normal brain dose. One thing I'll touch on, and I'll kind of piggyback on Terry's discussion about targets that are away from the isocenter. The one thing that we've considered is the tolerance on the ExacTrac system. Usually, there's I don't know, one degree, one-millimeter tolerance. And if you allow that one degree and one millimeter, essentially, if you look at something that says seven centimeters away and we put a one-degree rotation on that in addition to a one-millimeter translation, you're looking at probably close to two millimeters of offset. And this is just a two-dimensional representation.
So, what we've done is looked at our ExacTrac tolerances and based the margin recipe on that. And then there's a really good paper by Jay Chang out of New York who's done a pretty rigorous three-dimensional mathematical review of incorporating rotational setup uncertainty. So, what happens in the case where we've got targets that are far apart and the physician wants to use a smaller margin? Well, we consider either lowering our tolerance on what we're gonna allow on the actual verification images, or we might possibly break groups up and then shorten the distance from the isocenter to the actual targets. So, this is our experience thus far. As I mentioned, we've been going about seven months, a little over seven months. We started in December. There was a large Christmas break, and then we picked back up and as of July, two weeks ago, we've done about 67 patients. So, I'd say we average about two to three metastasis patients per week. And thus far, we've done about 139 total targets.
As far as statistics go, I'll give you the whole range here. We've treated things as small as 0.2 CCs all the way up to almost 50 CCs. The smallest target we've treated has been eight millimeters, and the largest target has had a maximum dimension of 6.5 centimeters. And this is kind of the results that we see with the Multiple Mets application. Really nice tight conformal dose distributions. This is 8.4-millimeter target. So, one of the challenges we want to look at these conformity indices and gradient indexes. And you can see the numbers I've presented here and it's across a range, but when you get two targets that are in close proximity to each other, a lot of these things, I don't think they have the same meaning. This single target over here, you could probably calculate a conformity index and look at the dose coverage. But when you think about the gradient index, it's gonna overlap with the near target. So, we have to take these indexes with a grain of salt when we think about it in a Multiple Mets environment.
So, real quick. Version 2.0. There are a lot of additional features. In Version 1.5, it's a very templated system. You've gone from doing things in hours and days to doing things in minutes. But with this system, they give you the capability to drill down and provide additional detail and editing capabilities. I'll just talk briefly about a couple of them. Specifically, the beam's-eye-view margin optimization, the dose constraints, and the jaw tracking. Are there a lot of eye-plain users that have done conformal arcs? I don't know if you recall, one of the tricks that you could do to increase the gradient index is to do a negative margin and just prescribe to a lower isodose line. They've incorporated this feature into the optimization of all targets. So it's maybe a little hard to see here, but you can see the leaves encroaching on the target here. And they're outside of it here. So, what that does is, if you look at these two distributions, you can see in this case where they've applied a negative margin, prescribed to a lower isodose line. They get a little bit better conformity to the target and a little bit better gradient index.
The organ at risk sparing. This is something that wasn't included in version 1.5. You're kind of at the mercy of what you chose in your template, but now they've incorporated some logic into the optimization so that you can actually start to pull dose away. In this case, if you look on the left, there's some excess dose spilling into the brainstem and on the right, they've been able to optimize it to come off. The other thing this is specific to the Elekta machine and taking advantage of the virtual one-millimeter MLC is that you can snap the jaw or bring the jaw past the edge of the MLC. So, now you can get a little bit better conformality on some of your smaller targets.
So, I had the opportunity to work with Brainlab last week. They gave me a sneak peek at some of the new software. I provided 10 patients to them and I think 40 targets anonymized. And to be honest, I gave them really, really bad patients. I didn't want to cherry-pick out the good stuff, I wanted to see what the new version of the software would do, and this is probably the thing that stood out to me the most. We looked at V12 as a normal brain marker. And if you look here across the top, these values in Gray are the original plans that we treated clinically. And then when Brainlab went back in the version two of the software, for those 10 patients using version two, and then version two with jaw tracking in yellow, you can see in some of these cases, we've had a rather significant drop in V12 dose. It's sometimes as high as 40%. So, I was pretty pleased to see that. Again, this is a small sample size, but I think the initial results are very promising.
They also provided some data on the conformity end index. Again, we can see that this average is a little bit different than the average before. This is a subset of 10 patients outside of my 67. And I gave them pretty crummy patients, to begin with, but there was an improvement. You can see here for some of these targets, there's not a huge improvement, it's pretty slight. The range here goes from one to 1.5. And most of the time, you're gonna see a conformity index in the range of 1.3 to 1.6. Same similar results for gradient index. The change is probably not statistically significant. But in the broader sense, it does appear to be a slight improvement for the subset of patients that we've seen. Again, this is a follow-up graph. Marginal improvement. We usually look for the gradient index to be somewhere around three to four.
And so in conclusion, I think this is a very accurate, precise, efficient system for us. We're a very busy practice. As I mentioned, we're doing about two to three patients per week. I don't know if anyone had the opportunity to attend the summer school last week, but George Ding had an interesting piece about the ultimate end-to-end test is when you can get follow-up imaging on your patient. And we've been able to do this with some of our patients and it's really reassuring to go in and see what you've been able to do with the system. I think that's it.
So, as you're gonna find out in a couple of slides, our program is relatively new. We've been going for about seven months. And anytime you bring up a new program, it's really important to have a good team behind you to work with. And I've been really fortunate to have that team of physicians, physicists, dosimetrists, therapists, and the rest of the team that I work with. I'd also like to give a shout-out to a couple of other people from Elekta Brainlab and SunNuclear who provided some information for this talk. So, our equipment, we have a Versa HD, and this is just a machine that we decided was gonna be mostly dedicated to stereotactic radiosurgery and SBRT. Therefore, we only have four energies. Four photon energies, we went with 6 and 10 and 6 and 10 flattening filter-free. There's a couple of philosophies behind that, but the first one for me as a busy physicist is a cut-down on commissioning time, but also cuts down on a lot of additional QA.
This particular machine comes with HexaPod 60 Couch from Elekta, obviously the Brainlab ExacTrac system, and the Agility MLC. So, the Agility MLC is kind of the heart of the delivery system. And for those that aren't familiar with this device, it's 165-millimeter leaf, and they are capable of spanning the entire 40x40 field. So, you are not limited by field size with this device. The transmission is very low at less than 0.4% and it is capable of what they call a virtual one-millimeter beamlet per aperture. So, the MLCs themselves can be positioned to better than one millimeter, and the jaws themselves can be positioned to better than one millimeter. So, we bring this program up, obviously, you go through the standard data collection process, the validation process of your model. Probably one of the more important things you can do is an end-to-end test. Brainlab does provide a really nice cranial phantom with a hidden target for you to do that. We did that test. In addition, we also were able to acquire another phantom and do a little bit of film. We looked at film obviously for its spatial resolution. Gafchromic EBT3 film, unfortunately, has a dose limitation of about 10 Gray. The good news is that I think it's Ashland Scientific now makes an extended dose range film that goes out to about 40 Gray. I think the bad news is it cost twice as much.
The software has pretty nicely configured the elements software so that you can map QA plans over to a phantom. And one of the things that we did was do point doses and we went with this SunNuclear StereoPHAN device. It's indexed to the couch. That kind of makes some of the measurements easy. And this one's milled for a small volume ion chamber. We've used a SunNuclear MapCHECK device to check fluences with the MapPHAN and had pretty good results. And then SunNuclear has just recently released their StereoPHAN, which is basically a 7.7-centimeter squared device with I believe it's 1,013 detectors. So you get a much higher detector density and increased spatial resolution. That device again fits inside their StereoPHAN.
As far as pretreatment quality assurance for single fraction stuff, we like to do a pretreatment once in a while and we look for that to be about 0.7 millimeters. We use the standard daily verification of better than one millimeter for all of our fractionated treatments. I think the real reason a lot of us are probably here are probably interested in the Multiple Mets application. I don't know if there's maybe a show in the audience how many people have seen this before, have actually done Multiple Mets, multiple isocenters. It's a very excruciating process. Time seems to slow down and things can take hours and days not only to plan this but also to deliver it depending on how much the patient can tolerate. One of the newer, I shouldn't say newer, but one of the more recent approaches is the VMAT approach. I'm sure more of us are familiar with that, single-isocenter. You optimize multiple targets at one time. I think it's a little bit quicker than the other approach. But one of the challenges you run into is something called the island blocking problem that was proposed by King.
When you have two targets that fall along the same path as the geometric delivery of the beam, the software is trying to optimize those two targets at exactly the same time. And what happens is you get an open spot in between where most likely there's either an organ at risk or a normal brain. So, you get this increase in normal brain dose. Obviously, you could work to try and geometrically avoid those approaches, but it gets more challenging, obviously, as you increase with the number of metastasis. Brainlab has taken a pretty unique approach with this, they take a dynamic conformal approach. You can treat Multiple Mets, you do it with a single-isocenter. And it's a dynamic conformal arc approach. But the way that they get around the island blocking program is that a given couch angle if there are gonna be two passes that hit the same or that hit two targets in one pass, they'll do it with two passes. So basically, the first pass, they hit the first target, the second pass, they hit the second target. And this significantly cuts down on a normal brain dose.
For us, the strategy and what we look at when we consider treating somebody with SRS or fractionated SRS, we look at the size. A lot of times post-op cavities will tend to be large, but things on the order of 3cm or larger, we look to fractionate depending on what they're next to. If it's it's next to an organ at risk, we'll fractionate. Targets especially in the Multiple Mets environment tend to have close proximity to each other causing dose bridging. And obviously, we look at normal brain dose. One thing I'll touch on, and I'll kind of piggyback on Terry's discussion about targets that are away from the isocenter. The one thing that we've considered is the tolerance on the ExacTrac system. Usually, there's I don't know, one degree, one-millimeter tolerance. And if you allow that one degree and one millimeter, essentially, if you look at something that says seven centimeters away and we put a one-degree rotation on that in addition to a one-millimeter translation, you're looking at probably close to two millimeters of offset. And this is just a two-dimensional representation.
So, what we've done is looked at our ExacTrac tolerances and based the margin recipe on that. And then there's a really good paper by Jay Chang out of New York who's done a pretty rigorous three-dimensional mathematical review of incorporating rotational setup uncertainty. So, what happens in the case where we've got targets that are far apart and the physician wants to use a smaller margin? Well, we consider either lowering our tolerance on what we're gonna allow on the actual verification images, or we might possibly break groups up and then shorten the distance from the isocenter to the actual targets. So, this is our experience thus far. As I mentioned, we've been going about seven months, a little over seven months. We started in December. There was a large Christmas break, and then we picked back up and as of July, two weeks ago, we've done about 67 patients. So, I'd say we average about two to three metastasis patients per week. And thus far, we've done about 139 total targets.
As far as statistics go, I'll give you the whole range here. We've treated things as small as 0.2 CCs all the way up to almost 50 CCs. The smallest target we've treated has been eight millimeters, and the largest target has had a maximum dimension of 6.5 centimeters. And this is kind of the results that we see with the Multiple Mets application. Really nice tight conformal dose distributions. This is 8.4-millimeter target. So, one of the challenges we want to look at these conformity indices and gradient indexes. And you can see the numbers I've presented here and it's across a range, but when you get two targets that are in close proximity to each other, a lot of these things, I don't think they have the same meaning. This single target over here, you could probably calculate a conformity index and look at the dose coverage. But when you think about the gradient index, it's gonna overlap with the near target. So, we have to take these indexes with a grain of salt when we think about it in a Multiple Mets environment.
So, real quick. Version 2.0. There are a lot of additional features. In Version 1.5, it's a very templated system. You've gone from doing things in hours and days to doing things in minutes. But with this system, they give you the capability to drill down and provide additional detail and editing capabilities. I'll just talk briefly about a couple of them. Specifically, the beam's-eye-view margin optimization, the dose constraints, and the jaw tracking. Are there a lot of eye-plain users that have done conformal arcs? I don't know if you recall, one of the tricks that you could do to increase the gradient index is to do a negative margin and just prescribe to a lower isodose line. They've incorporated this feature into the optimization of all targets. So it's maybe a little hard to see here, but you can see the leaves encroaching on the target here. And they're outside of it here. So, what that does is, if you look at these two distributions, you can see in this case where they've applied a negative margin, prescribed to a lower isodose line. They get a little bit better conformity to the target and a little bit better gradient index.
The organ at risk sparing. This is something that wasn't included in version 1.5. You're kind of at the mercy of what you chose in your template, but now they've incorporated some logic into the optimization so that you can actually start to pull dose away. In this case, if you look on the left, there's some excess dose spilling into the brainstem and on the right, they've been able to optimize it to come off. The other thing this is specific to the Elekta machine and taking advantage of the virtual one-millimeter MLC is that you can snap the jaw or bring the jaw past the edge of the MLC. So, now you can get a little bit better conformality on some of your smaller targets.
So, I had the opportunity to work with Brainlab last week. They gave me a sneak peek at some of the new software. I provided 10 patients to them and I think 40 targets anonymized. And to be honest, I gave them really, really bad patients. I didn't want to cherry-pick out the good stuff, I wanted to see what the new version of the software would do, and this is probably the thing that stood out to me the most. We looked at V12 as a normal brain marker. And if you look here across the top, these values in Gray are the original plans that we treated clinically. And then when Brainlab went back in the version two of the software, for those 10 patients using version two, and then version two with jaw tracking in yellow, you can see in some of these cases, we've had a rather significant drop in V12 dose. It's sometimes as high as 40%. So, I was pretty pleased to see that. Again, this is a small sample size, but I think the initial results are very promising.
They also provided some data on the conformity end index. Again, we can see that this average is a little bit different than the average before. This is a subset of 10 patients outside of my 67. And I gave them pretty crummy patients, to begin with, but there was an improvement. You can see here for some of these targets, there's not a huge improvement, it's pretty slight. The range here goes from one to 1.5. And most of the time, you're gonna see a conformity index in the range of 1.3 to 1.6. Same similar results for gradient index. The change is probably not statistically significant. But in the broader sense, it does appear to be a slight improvement for the subset of patients that we've seen. Again, this is a follow-up graph. Marginal improvement. We usually look for the gradient index to be somewhere around three to four.
And so in conclusion, I think this is a very accurate, precise, efficient system for us. We're a very busy practice. As I mentioned, we're doing about two to three patients per week. I don't know if anyone had the opportunity to attend the summer school last week, but George Ding had an interesting piece about the ultimate end-to-end test is when you can get follow-up imaging on your patient. And we've been able to do this with some of our patients and it's really reassuring to go in and see what you've been able to do with the system. I think that's it.