Transcript
Thank you, Dr. [inaudible 00:00:03.500], and thank you Brainlab for inviting me out to Germany for this talk. This is very exciting. I think it's great to have so many people from different countries in one meeting. It's just fascinating. So I'm gonna talk about MLC Leakage Analysis for Quality Assurance. Again, my name is Charles Geraghty. I work at Anne Arundel Medical Center which is in Annapolis, Maryland. And here's my disclosure.
Just to tell you a little bit about my center, it is a large community hospital, was founded in 1902. We have 400 beds in the main hospital building. We have a number of locations in Maryland. In the radiation oncology department, we have three LINACS. We have 21EX with an OBI. We have a Novalis TX with Exactrac. We treat [inaudible 00:00:56.225] for our multiple mets treatments, and we have TrueBeam with OSMS and Calypso.
Our special procedures range from cranial SRS and SRT, lung, spine, pancreas, SBRT. We do HDR vaginal cylinder and partial breast irradiation with SAVI and Contura. We do LDR prostate seed implants. And in addition to our Brainlab planning systems with the Iplan and the elements, we also have Eclipse with AAA and Acuros.
Our staff is interesting to compare to those in Europe. We have two geneticists on staff. We have two social workers, two patient financial advisors. In the United States, one of the hot topics is financial toxicity which is the detriment, the financial burden of the treatments on the patients.
We also have smoking cessation staff which work to encourage people in the community to quit smoking to reduce the incidence of lung cancer. We have a speech therapist. We have one medical physics resident.
Our experience is a community hospital experience. This data is as of July of this year. We had treated seven patients with our multiple mets technique. That number is up to eight or nine, I believe now, or possibly more. Our average number of targets treated per multiple mets treatment is three and ranges from two to five. Average [inaudible 00:02:36.573] target is 1.34. Average prescription dose is 22 Gray. We use a 0.5 millimeters half-degree tolerance which I'll also talk about is very important for the treatment accuracy. And of course, we use a six degree of freedom couch with Exactrac which I think is essential for multiple metstechnique because of that angular uncertainty in the treatment.
We do have historical ties to Brainlab at Anne Arundel. Even though our center is a community hospital, it is relatively smaller compared to the large institutions in the United States. If you look in the Brainlab Museum, which is very interesting, you'll see something there about BrainScan and early development of the BrainScan system. And if you read that second paragraph in there above the BrainScan picture, I'll just read this for you, it says that Stefan Vilsmeier became fascinated with stereotactic radiosurgery in July 1989 when Bob Siddon, a physicist at Brigham and Women's Hospital in Boston, demonstrated his homegrown radiosurgery system. It would inspire Brainlab to develop its own stereotactic radiosurgery software with a sophisticated graphical interface which was BrainSCAN.
And that physicist who inspired Stefan, Bob Siddon, after working at Brigham Women's came to Anne Arundel Medical Center. He established our stereotactic radiosurgery program in cooperation with neurosurgery, radiation oncology, of course, but he was very influential at my particular site. So we do have a historical tie to Brainlab, which is special in my facility.
The recent emergence of multiple mets techniques has been studied at numerous facilities. This is a United States perspective. University of Alabama Birmingham, Henry Ford Cancer Center, UCLA, and Duke have been some of the leaders on the United States side investigating different techniques for multiple mets including VMAT versus dynamic conformal arc. I just have a selection of papers from the United States.
And when we went to commission our own multiple mets [inaudible 00:04:52.099] our commissioning measurements was not comprehensive. Our commissioning measurements included SFD diode measurements, MapCheck, ArcCheck, and Gafchromic film measurements.
Our measured data is [inaudible 00:05:07] treatment planning system and here's just a small portion of our commissioning data. One of the things we looked at was on-axis and off-axis small fields and the accuracy of the planning system to duplicate what we measured. We looked at it again for elements of pencil beam algorithm versus AAA and we also looked at our RadCalc and we found that the results were quite good even going down to small fields. And in this particular data set it went down to 6 millimeters by 6 millimeters. One of the interesting things we'll find, I think, looking at multiple mets treatment is how small did we go. How small should we allow that MLC aperture to go?
And here's one study from Emory University in Atlanta, Georgia where they investigated the dosimetric effect of patient rotations and translations on the plan. And this is somewhere to what our previous individual presented just a moment ago. They found that 0.5-degree rotational error, the D95 and V95 coverage rates were greater than or equal to 95% in all cases. Even going out to 0.8 centimeters off the axis on a PTV, which is very large. This was off-axis.
And again, at 2 degrees they found that the D95 and V95 values were only greater than 95% for 63% of the targets. So then at 2 degrees, you have a very large dosimetric effect.
And again, I just want to emphasize the importance of MLC leakage and other MLC features in this delivery. So here's a typical beam aperture in a multiple mets field where we can see that the... So this is the collimator, the square collimator field size, with a collimator jaws, and our [inaudible 00:07:10.089] coming in here blocking the field. Most of the field is actually blocked MLC which is very different from our traditional SRS delivery where we would normally have just, if we were using MLCs, one MLC aperture on the central axis.
So now, with this treatment, I believe the MLC leakage is more of a physics consideration than a typical MLC-based LINAC SRS treatment. And also, because we're treating multiple metastases at the same time, in these kind of cases those metastases can sometimes be quite small and whereas before how we would normally treat that is we would use cone-based treatments. Our largest cone size in my facility is 1.5 centimeters in diameter. So anything 1.5 centimeters or smaller, and it could be smaller than that, we would conventional treat with cones.
Now, with our multiple mets technique, we're using MLCs to treat those patients. And also, I think that's an important physics consideration is how, you know, if we're doing this at the facility just to make sure that when we're going down that small that dosimetrically we're very accurate in a small field-size. In this talk, we emphasize low dose. Our main inspiration with this was what was the effect of this MLC-blocked field. Is this an accurate...are we modeling this accurately in the planning system?
And in version 1.5 multiple brain mets SRS it offers a pencil beam algorithm for all calculation types and Monte Carlo for RTQA. Pencil beam algorithm has an MLC leakage parameter that can be modified in Physics Administration. The parameter incorporates both intra and interleaf leakage and is measured as the ratio of dose on the central axis from the 10 by 10 field with blocked MLCs to an open 10 by 10 field. And let me see, that is the parameter here. And those of us who are physicists here, and I hope there are also non-physicists here in this doc, but you can see we chose a parameter of 1.4%. I'll get into that when we measure this on the machine. We actually measured 1.1%, but on purpose, we adjusted it to be different from what we had measured to try to increase the planning accuracy.
MLC leakage parameters in Physics Administration are used for pencil beam calculation only. The Monte Carlo algorithm simulates the MLCs including MLC geometry and material properties specific to the MLC type as part of the calculation. So this is one reason why in the forthcoming versions of the software that the Monte Carlo algorithm could be appealing, could more accurately model for MLCs.
One of the first measurements we made to verify the [inaudible 00:10:07.537] plans was with ArcCheck. We created plans from two to eight targets. First with the 1.1% leakage parameter. When we measured our plans what we found was that there were a number of points in the treatment field where our measuring dose was higher than the treatment plan system expected. So we adjusted our leakage parameter up thinking that maybe this will increase our accuracy.
So for [inaudible 00:10:39.240] here's the case. This was the worst one we had. It was targeted for, let's see, how many lesions. It's numerous lesions. The couch kicks have all been set to zero because the clearance of the ArcCheck phantom. And you can see here we have kind of a low bath of points that are measuring higher than what we expected. And again, this was the worst plan that we saw with a passing rate of 3%, 3 millimeters was 61.5%.
After the adjusted leakage parameter of 1.4%, we saw much better ArcCheck results. And then we considered, well, is this just a feature of ArcCheck with the diodes? Is there something going on here? Let's confirm this with other measurement techniques.
So we did gafchromic film measurements. Plans were measured EBT3 film in solid water. The couch angles were projected to zero similar to our ArcCheck measurement. We scanned the films with an Epson 12000XL with a transparency adapter and we analyzed results in the RIT software using a red channel. And I put a reference here, we did not use their exact technique. This is a technique from Henry Ford. I find it to be a very fascinating paper where they really took their film dosimetry to a higher level and that was kind of a inspiration for us.
Here's an example of a plan measure with our gafchromic film. You can see there's very good green between what we measured and what was calculated in the planning system. And we're looking in this low dose region was the area that was of particular interest to us because most of the points that are measured in the ArcCheck phantom are not actual points where the phantom is getting full dose. It's mostly just a low-dose bath of the leakage.
And we also used a farmer chamber, a simple measurement. We put farmer chamber in a solid water phantom, delivered the plans with the same couch kicks to zero, the same technique as the previous methods. And we compared the dose at isocenter, measured the farmer chamber to the treatment planning system dose. Here's one plan. You can see that, you know, these targets are getting full dose and the isocenter is getting a lower dose.
We did this for five plans with four patients and we found that a range of doses at isocenter from 1.8 to 3.22 Gray and the difference between the measured and the planning system dose range from -0.15 to 0.16 Gray with an average point -0.02 Gray.
So that confirms...that made us feel good that [inaudible00:13:33.492] leakage parameter we didn't do something wrong that we were validated.
In Brainlab multiple Mets planning version 2.0, Brainlab will offer modeling changes for clinical patient calculations including Monte Carlo for clinical patient calculations and a change in the pencil beam modeling. And I'll just read this quote from Brainlab.
"For BrainMets 2.0 we modified the pencil beam in a way that we allow for a second Source Function Correction. The SFC is the parameter that is adjusted during beam data processing. The second SFC allows for a more accurate fitting of the penumbra and the peaks and valleys in the Transversal Profile Shape. At the end, using the second source function, the penumbra and the leakage are modeled better. It does not require new measurements."
So now they're modifying the pencil beam algorithm to not only fit the peak where we're delivering most of the dose but the valley where we are...and the penumbras where we are delivering low dose.
And here is some screen capture from version 2.0. You could see that there is, on the side, there's a Monte Carlo button, I believe. And it shows pencil-beam calculation, a Monte Carlo recalculation, and the Monte Carlo optimization.
And some physics considerations, again, are the small MLC fields going down to 5 millimeters by 5 millimeters. And we are traveling, starting to travel this road and we are going to look at our peak profiles at these smaller MLCs apertures. And other thing to think of is what are the expected deviations between pencil beam and Monte Carlo with small off-axis MLC fields? That's just a question for all the physicists out there.
So, in conclusion, single isocenter multi-target SRS technique was implemented clinically at our community cancer center in 2017. MLC leakage parameter for pencil beam calculations was adjusted to 1.4% and confirmed with QA measurements. And multiple mets planning in version 2.0 will feature Monte Carlo for patient calculations and an updated pencil beam model. Thank you very much.
Just to tell you a little bit about my center, it is a large community hospital, was founded in 1902. We have 400 beds in the main hospital building. We have a number of locations in Maryland. In the radiation oncology department, we have three LINACS. We have 21EX with an OBI. We have a Novalis TX with Exactrac. We treat [inaudible 00:00:56.225] for our multiple mets treatments, and we have TrueBeam with OSMS and Calypso.
Our special procedures range from cranial SRS and SRT, lung, spine, pancreas, SBRT. We do HDR vaginal cylinder and partial breast irradiation with SAVI and Contura. We do LDR prostate seed implants. And in addition to our Brainlab planning systems with the Iplan and the elements, we also have Eclipse with AAA and Acuros.
Our staff is interesting to compare to those in Europe. We have two geneticists on staff. We have two social workers, two patient financial advisors. In the United States, one of the hot topics is financial toxicity which is the detriment, the financial burden of the treatments on the patients.
We also have smoking cessation staff which work to encourage people in the community to quit smoking to reduce the incidence of lung cancer. We have a speech therapist. We have one medical physics resident.
Our experience is a community hospital experience. This data is as of July of this year. We had treated seven patients with our multiple mets technique. That number is up to eight or nine, I believe now, or possibly more. Our average number of targets treated per multiple mets treatment is three and ranges from two to five. Average [inaudible 00:02:36.573] target is 1.34. Average prescription dose is 22 Gray. We use a 0.5 millimeters half-degree tolerance which I'll also talk about is very important for the treatment accuracy. And of course, we use a six degree of freedom couch with Exactrac which I think is essential for multiple metstechnique because of that angular uncertainty in the treatment.
We do have historical ties to Brainlab at Anne Arundel. Even though our center is a community hospital, it is relatively smaller compared to the large institutions in the United States. If you look in the Brainlab Museum, which is very interesting, you'll see something there about BrainScan and early development of the BrainScan system. And if you read that second paragraph in there above the BrainScan picture, I'll just read this for you, it says that Stefan Vilsmeier became fascinated with stereotactic radiosurgery in July 1989 when Bob Siddon, a physicist at Brigham and Women's Hospital in Boston, demonstrated his homegrown radiosurgery system. It would inspire Brainlab to develop its own stereotactic radiosurgery software with a sophisticated graphical interface which was BrainSCAN.
And that physicist who inspired Stefan, Bob Siddon, after working at Brigham Women's came to Anne Arundel Medical Center. He established our stereotactic radiosurgery program in cooperation with neurosurgery, radiation oncology, of course, but he was very influential at my particular site. So we do have a historical tie to Brainlab, which is special in my facility.
The recent emergence of multiple mets techniques has been studied at numerous facilities. This is a United States perspective. University of Alabama Birmingham, Henry Ford Cancer Center, UCLA, and Duke have been some of the leaders on the United States side investigating different techniques for multiple mets including VMAT versus dynamic conformal arc. I just have a selection of papers from the United States.
And when we went to commission our own multiple mets [inaudible 00:04:52.099] our commissioning measurements was not comprehensive. Our commissioning measurements included SFD diode measurements, MapCheck, ArcCheck, and Gafchromic film measurements.
Our measured data is [inaudible 00:05:07] treatment planning system and here's just a small portion of our commissioning data. One of the things we looked at was on-axis and off-axis small fields and the accuracy of the planning system to duplicate what we measured. We looked at it again for elements of pencil beam algorithm versus AAA and we also looked at our RadCalc and we found that the results were quite good even going down to small fields. And in this particular data set it went down to 6 millimeters by 6 millimeters. One of the interesting things we'll find, I think, looking at multiple mets treatment is how small did we go. How small should we allow that MLC aperture to go?
And here's one study from Emory University in Atlanta, Georgia where they investigated the dosimetric effect of patient rotations and translations on the plan. And this is somewhere to what our previous individual presented just a moment ago. They found that 0.5-degree rotational error, the D95 and V95 coverage rates were greater than or equal to 95% in all cases. Even going out to 0.8 centimeters off the axis on a PTV, which is very large. This was off-axis.
And again, at 2 degrees they found that the D95 and V95 values were only greater than 95% for 63% of the targets. So then at 2 degrees, you have a very large dosimetric effect.
And again, I just want to emphasize the importance of MLC leakage and other MLC features in this delivery. So here's a typical beam aperture in a multiple mets field where we can see that the... So this is the collimator, the square collimator field size, with a collimator jaws, and our [inaudible 00:07:10.089] coming in here blocking the field. Most of the field is actually blocked MLC which is very different from our traditional SRS delivery where we would normally have just, if we were using MLCs, one MLC aperture on the central axis.
So now, with this treatment, I believe the MLC leakage is more of a physics consideration than a typical MLC-based LINAC SRS treatment. And also, because we're treating multiple metastases at the same time, in these kind of cases those metastases can sometimes be quite small and whereas before how we would normally treat that is we would use cone-based treatments. Our largest cone size in my facility is 1.5 centimeters in diameter. So anything 1.5 centimeters or smaller, and it could be smaller than that, we would conventional treat with cones.
Now, with our multiple mets technique, we're using MLCs to treat those patients. And also, I think that's an important physics consideration is how, you know, if we're doing this at the facility just to make sure that when we're going down that small that dosimetrically we're very accurate in a small field-size. In this talk, we emphasize low dose. Our main inspiration with this was what was the effect of this MLC-blocked field. Is this an accurate...are we modeling this accurately in the planning system?
And in version 1.5 multiple brain mets SRS it offers a pencil beam algorithm for all calculation types and Monte Carlo for RTQA. Pencil beam algorithm has an MLC leakage parameter that can be modified in Physics Administration. The parameter incorporates both intra and interleaf leakage and is measured as the ratio of dose on the central axis from the 10 by 10 field with blocked MLCs to an open 10 by 10 field. And let me see, that is the parameter here. And those of us who are physicists here, and I hope there are also non-physicists here in this doc, but you can see we chose a parameter of 1.4%. I'll get into that when we measure this on the machine. We actually measured 1.1%, but on purpose, we adjusted it to be different from what we had measured to try to increase the planning accuracy.
MLC leakage parameters in Physics Administration are used for pencil beam calculation only. The Monte Carlo algorithm simulates the MLCs including MLC geometry and material properties specific to the MLC type as part of the calculation. So this is one reason why in the forthcoming versions of the software that the Monte Carlo algorithm could be appealing, could more accurately model for MLCs.
One of the first measurements we made to verify the [inaudible 00:10:07.537] plans was with ArcCheck. We created plans from two to eight targets. First with the 1.1% leakage parameter. When we measured our plans what we found was that there were a number of points in the treatment field where our measuring dose was higher than the treatment plan system expected. So we adjusted our leakage parameter up thinking that maybe this will increase our accuracy.
So for [inaudible 00:10:39.240] here's the case. This was the worst one we had. It was targeted for, let's see, how many lesions. It's numerous lesions. The couch kicks have all been set to zero because the clearance of the ArcCheck phantom. And you can see here we have kind of a low bath of points that are measuring higher than what we expected. And again, this was the worst plan that we saw with a passing rate of 3%, 3 millimeters was 61.5%.
After the adjusted leakage parameter of 1.4%, we saw much better ArcCheck results. And then we considered, well, is this just a feature of ArcCheck with the diodes? Is there something going on here? Let's confirm this with other measurement techniques.
So we did gafchromic film measurements. Plans were measured EBT3 film in solid water. The couch angles were projected to zero similar to our ArcCheck measurement. We scanned the films with an Epson 12000XL with a transparency adapter and we analyzed results in the RIT software using a red channel. And I put a reference here, we did not use their exact technique. This is a technique from Henry Ford. I find it to be a very fascinating paper where they really took their film dosimetry to a higher level and that was kind of a inspiration for us.
Here's an example of a plan measure with our gafchromic film. You can see there's very good green between what we measured and what was calculated in the planning system. And we're looking in this low dose region was the area that was of particular interest to us because most of the points that are measured in the ArcCheck phantom are not actual points where the phantom is getting full dose. It's mostly just a low-dose bath of the leakage.
And we also used a farmer chamber, a simple measurement. We put farmer chamber in a solid water phantom, delivered the plans with the same couch kicks to zero, the same technique as the previous methods. And we compared the dose at isocenter, measured the farmer chamber to the treatment planning system dose. Here's one plan. You can see that, you know, these targets are getting full dose and the isocenter is getting a lower dose.
We did this for five plans with four patients and we found that a range of doses at isocenter from 1.8 to 3.22 Gray and the difference between the measured and the planning system dose range from -0.15 to 0.16 Gray with an average point -0.02 Gray.
So that confirms...that made us feel good that [inaudible00:13:33.492] leakage parameter we didn't do something wrong that we were validated.
In Brainlab multiple Mets planning version 2.0, Brainlab will offer modeling changes for clinical patient calculations including Monte Carlo for clinical patient calculations and a change in the pencil beam modeling. And I'll just read this quote from Brainlab.
"For BrainMets 2.0 we modified the pencil beam in a way that we allow for a second Source Function Correction. The SFC is the parameter that is adjusted during beam data processing. The second SFC allows for a more accurate fitting of the penumbra and the peaks and valleys in the Transversal Profile Shape. At the end, using the second source function, the penumbra and the leakage are modeled better. It does not require new measurements."
So now they're modifying the pencil beam algorithm to not only fit the peak where we're delivering most of the dose but the valley where we are...and the penumbras where we are delivering low dose.
And here is some screen capture from version 2.0. You could see that there is, on the side, there's a Monte Carlo button, I believe. And it shows pencil-beam calculation, a Monte Carlo recalculation, and the Monte Carlo optimization.
And some physics considerations, again, are the small MLC fields going down to 5 millimeters by 5 millimeters. And we are traveling, starting to travel this road and we are going to look at our peak profiles at these smaller MLCs apertures. And other thing to think of is what are the expected deviations between pencil beam and Monte Carlo with small off-axis MLC fields? That's just a question for all the physicists out there.
So, in conclusion, single isocenter multi-target SRS technique was implemented clinically at our community cancer center in 2017. MLC leakage parameter for pencil beam calculations was adjusted to 1.4% and confirmed with QA measurements. And multiple mets planning in version 2.0 will feature Monte Carlo for patient calculations and an updated pencil beam model. Thank you very much.