Transcript
Good afternoon, everybody. I'll spend the next 15 minutes or so with you to share our experience with Spine Elements. I'd like to start by thanking some of my colleagues. Daniel Saentz is one of our faculty members and a former resident. He did a lot of the work together with George and Rodrigo, they were graduate students. Rick Crownover is one of our radiation oncologists that's helping us with the spine SBRT. And also acknowledge RtSafe for the [inaudible 00:00:33] they provided to us to do the [inaudible 00:00:34] part of this project. And also Brainlab, more specifically Bogdan, Rebecca, Corey, and the bunch of engineering folks that helped us deliver the implementation of [inaudible 00:00:44] in its alpha and beta phase.
In terms of an outline, I'll speak a little bit in terms of Element's introduction and we will do an overview. I'll share with you the treatment planning study that we did, and also the commissioning and the testing that we performed prior to going clinical. And we'll share some conclusions.
So, the history of Elements at the University of Texas in San Antonio started back in 2015. We had a research version of early prototypes, but we actually went clinical at the Mays Cancer Center in 2017. We started with the commission and acceptance testing of the clinical version of the software, and then the first phase of implementation was specific to the brain mets Elements, and more recently we implemented the spine SRS as BRT.
We looked at the quick overview of the spine. This is an application that fits into the environment of Elements. As you know, it's elaborate and has many different functional elements that contribute to planning, fusion, contouring, evaluation of a plan. And it is designed to optimize conformal dose distributions for the spinal anatomy. Three of its more significant components are the spinal curvature correction, the spine contouring, and the spine VMAT optimization.
For the first one, it's mainly the spinal curvature correction. It's a very important one and a novel introduction of that functionality came with the Brainlab Elements. And that's because, with the MR images, we do use them to delineate the spinal targets, particularly in the re-treatment setting. So, it's important to be able to visualize them clearly. But in doing so, we bring an MR dataset that is typically done in a slightly different position and a different time than the CT simulation was performed. So, when we fuse the two it's important to do that correctly. So, the spinal curvature correction is a new feature that's embedded in Elements that allows you to have a deformable registration between the CT and the MR images for the patient, where the bone is not deformed because it cannot be, whereas the tissue around the vertebrae is deformed. It allows you to map the two datasets in the correct anatomical position and capitalize on the contouring visualization with the target from the MRI and map it on the CT.
The spine contouring is also automated, and the beauty of it here is that as you draw the GTV, the system automatically draws the CTV. And that's based on the RTG, the 0631, or the surgical contouring guidelines for the spine. So, if the tumor is in the [inaudible 00:03:58] we automatically convert the tumor body in the pedicles, and if it's in the posterior aspect it will be able to contour automatically for the CTV the posterior process of the spine.
It also has a very smart VMAT optimization, and that's very important for the spine because it does have a very complex shape, and so we have shape exhibiting concavity. Traditional optimization struggles to carve out the dose from regions such as the spinal cord, which is [inaudible 00:04:34] in the spinal treatments. And Elements has implemented a very sophisticated way with splitting the VMAT, the general beams, and segmenting them in four different regions to accomplish that.
Let's look at a treatment study where we compared the three different planning systems, mainly the Elements against the Pinnacle and Monte Carlo. So, we'll do the Pinnacle and Monaco because this is what we have in our clinic. Pinnacle is a primary planning system that we use [inaudible 00:05:05] position, dose calculation [inaudible 00:05:07], and Monaco we use with the Monte Carlo dose calculation algorithm. So, what we did in the study, we took 10 patients with spinal lesions, we outlined the CTVs. We planned 20Gy using arc splitting, and we applied exact same geometry in Pinnacle and Monaco, exact same in the one that was derived from Elements, which was our primary plan [inaudible 00:05:33]. We had dose constraints to the PDV, D5 less than 25Gy. Spinal cord 0.035cc less than 14Gy. And as far as optimization strategy, we asked to cover 95% of the volume for all the plans. And the spinal cord sparing and dose falloff were pushed until the conformity or hot spot over the plan suffered.
For the assessment, we looked at plan qualities, namely the conformity index. The dose gradient index, the dose [inaudible 00:06:11] volume, and the physical dose gradient between target and spinal cord. That was a distance between the 20 and the 10Gy isodose line for the same slice. This is a slice showing for 3 of the 10 patients, and the 3 planning systems isodose distributions. Qualitatively, when you look at them, they look very similar, so we're able to obtain relatively equivalent distributions, as you can see here in those three pages. That was true for all 10 of them. But more specifically we want to look at the deviations, we start becoming a bit more quantitative in our comparisons. For those three patients you see here, we see the cord and the [inaudible 00:06:57] two separated families of the curves and deviations. What's not over here is that Elements did better in all the patients in terms of the deviations of the cord. Whereas the CTV coverage was about the same for all three patients shown here. But for the population of the 10, I believe that was also true.
Looking at more quantitatively, the comparison. The monitor units, the PTV coverage, and the PTV D5 were similar in all three planning environments. What's significantly different is the spinal cord maximum dose, the gradient index, the conformity index, and the distance from the 20 to the 10Gy line that was the gradient that we had found earlier. And here, we found that Elements did a better job in comparison with Pinnacle and Monaco.
So, in the conclusions of the planning study we found that a gradient index and spinal cord maximum dose was significantly smaller in Elements than the other planning systems with a p [inaudible 00:08:05] less than .01. Arc splitting does result in improved spinal cord sparing and improved dose gradient index, and the physical dose gradient is the smallest in Elements, and the smallest number was about 0.6 millimeters.
For the commissioning part, we followed the AAPM Medical Physics guidelines, MPPG 5.a publication that has a number of tests that one should perform to commission the planning system. And what you see here is screenshots of three of the analyses were small, off-axis, and large fields. And the graphs are a little small, but at full size they again are small themselves. So for a small field size, we compared the measured dose with the calculated dose using the Octavius 1000, that's an SRS dedicated 2d dosimeter. And the spacing between the measurement points are 2.5 millimeter center to center.
So, we got a very good agreement for the small field. Same was true for the larger field, which was about an eight by eight field. And also for the off-axis field, that was about four centimeters off-axis as well. So, in conclusion of that part of the commissioning, we found the 2d gamma analysis had a better than 90% passing rates for all points that we evaluated with 2%/2 millimeter evaluation criteria.
We also got the IROC spine Phantom that we irradiated after developing the plan for Elements for this anatomy. And we delivered 6Gy with a maximum dose of less than 5.25 as prescribed by IROC to the cord. And this is here the results that we obtained back from IROC. Also, a high passing score that shows the match between the Elements calculation and the film, the film being the blue, and the purple being the Elements, respectively. So, we felt pretty confident that the commissioning had gone well. We continued by doing some patient-specific QA. At least one of the patients that we [inaudible 00:10:34] using the SRS 1000 Phantom. Analyzed with global gamma relative to maximum dose in volume, 3%/1 millimeter in this case, and passing rates were over 95%. As you can appreciate, qualitatively, at least from the graphs [inaudible 00:10:52] shown here for the patient using the slides.
We also used another Phantom, we used a Delta4. We did plan in Elements, and then we mapped the Delta4, we calculated, and did the analysis as well. And yet another Phantom, the one from RTsafe, this is a true replica of one of the patients that we treated that was made using a 3D printer. We filled the [inaudible 00:11:25] here with water. We had an ion chamber of A16 in the spine and inside the tumor. So, when we did the analysis for that, these are the results that we received for the Delta4 comparison. Also a very high passing rate of 3%/1 millimeter, over 95%. And at the bottom, you see the difference that we measure in the CTV in the spinal cord between the plan and the measured values, also in very good agreement. So, we've been pretty pleased with the agreement that we've been receiving.
So, in conclusion, the Brainlab Spine Elements is uniquely suited for spine SRS from our experience. It has several unique, novel features that take advantage of the spine anatomy and it's specific to the spine curvature correction, automatic target delineation, optimization for complex spine target shapes, and intuitive planning tools that greatly facilitate automation. And also Monte Carlo calculation was important, especially in the presence of bone and for regions close to lines where you have two levels of inhomogeneity to account for.
Spine Elements generated plans, produced statistically significant better cord sparing and target conformity when prepared in a planning study with the Monte Carlo of Monaco planning system and the Pinnacle. And for the commissioning, we conducted with a combination of beam model verification, according to [inaudible 00:13:06] guidelines, the IROC practice accreditation, Phantom, the RTsafe, and patient-specific measurements, and we found that we had very good results. And since then, we have treated several patients very successfully with the same type of QA as shown here, with high confidence that we deliver exactly what we plan. And with that, I'd like to thank you for your time.
In terms of an outline, I'll speak a little bit in terms of Element's introduction and we will do an overview. I'll share with you the treatment planning study that we did, and also the commissioning and the testing that we performed prior to going clinical. And we'll share some conclusions.
So, the history of Elements at the University of Texas in San Antonio started back in 2015. We had a research version of early prototypes, but we actually went clinical at the Mays Cancer Center in 2017. We started with the commission and acceptance testing of the clinical version of the software, and then the first phase of implementation was specific to the brain mets Elements, and more recently we implemented the spine SRS as BRT.
We looked at the quick overview of the spine. This is an application that fits into the environment of Elements. As you know, it's elaborate and has many different functional elements that contribute to planning, fusion, contouring, evaluation of a plan. And it is designed to optimize conformal dose distributions for the spinal anatomy. Three of its more significant components are the spinal curvature correction, the spine contouring, and the spine VMAT optimization.
For the first one, it's mainly the spinal curvature correction. It's a very important one and a novel introduction of that functionality came with the Brainlab Elements. And that's because, with the MR images, we do use them to delineate the spinal targets, particularly in the re-treatment setting. So, it's important to be able to visualize them clearly. But in doing so, we bring an MR dataset that is typically done in a slightly different position and a different time than the CT simulation was performed. So, when we fuse the two it's important to do that correctly. So, the spinal curvature correction is a new feature that's embedded in Elements that allows you to have a deformable registration between the CT and the MR images for the patient, where the bone is not deformed because it cannot be, whereas the tissue around the vertebrae is deformed. It allows you to map the two datasets in the correct anatomical position and capitalize on the contouring visualization with the target from the MRI and map it on the CT.
The spine contouring is also automated, and the beauty of it here is that as you draw the GTV, the system automatically draws the CTV. And that's based on the RTG, the 0631, or the surgical contouring guidelines for the spine. So, if the tumor is in the [inaudible 00:03:58] we automatically convert the tumor body in the pedicles, and if it's in the posterior aspect it will be able to contour automatically for the CTV the posterior process of the spine.
It also has a very smart VMAT optimization, and that's very important for the spine because it does have a very complex shape, and so we have shape exhibiting concavity. Traditional optimization struggles to carve out the dose from regions such as the spinal cord, which is [inaudible 00:04:34] in the spinal treatments. And Elements has implemented a very sophisticated way with splitting the VMAT, the general beams, and segmenting them in four different regions to accomplish that.
Let's look at a treatment study where we compared the three different planning systems, mainly the Elements against the Pinnacle and Monte Carlo. So, we'll do the Pinnacle and Monaco because this is what we have in our clinic. Pinnacle is a primary planning system that we use [inaudible 00:05:05] position, dose calculation [inaudible 00:05:07], and Monaco we use with the Monte Carlo dose calculation algorithm. So, what we did in the study, we took 10 patients with spinal lesions, we outlined the CTVs. We planned 20Gy using arc splitting, and we applied exact same geometry in Pinnacle and Monaco, exact same in the one that was derived from Elements, which was our primary plan [inaudible 00:05:33]. We had dose constraints to the PDV, D5 less than 25Gy. Spinal cord 0.035cc less than 14Gy. And as far as optimization strategy, we asked to cover 95% of the volume for all the plans. And the spinal cord sparing and dose falloff were pushed until the conformity or hot spot over the plan suffered.
For the assessment, we looked at plan qualities, namely the conformity index. The dose gradient index, the dose [inaudible 00:06:11] volume, and the physical dose gradient between target and spinal cord. That was a distance between the 20 and the 10Gy isodose line for the same slice. This is a slice showing for 3 of the 10 patients, and the 3 planning systems isodose distributions. Qualitatively, when you look at them, they look very similar, so we're able to obtain relatively equivalent distributions, as you can see here in those three pages. That was true for all 10 of them. But more specifically we want to look at the deviations, we start becoming a bit more quantitative in our comparisons. For those three patients you see here, we see the cord and the [inaudible 00:06:57] two separated families of the curves and deviations. What's not over here is that Elements did better in all the patients in terms of the deviations of the cord. Whereas the CTV coverage was about the same for all three patients shown here. But for the population of the 10, I believe that was also true.
Looking at more quantitatively, the comparison. The monitor units, the PTV coverage, and the PTV D5 were similar in all three planning environments. What's significantly different is the spinal cord maximum dose, the gradient index, the conformity index, and the distance from the 20 to the 10Gy line that was the gradient that we had found earlier. And here, we found that Elements did a better job in comparison with Pinnacle and Monaco.
So, in the conclusions of the planning study we found that a gradient index and spinal cord maximum dose was significantly smaller in Elements than the other planning systems with a p [inaudible 00:08:05] less than .01. Arc splitting does result in improved spinal cord sparing and improved dose gradient index, and the physical dose gradient is the smallest in Elements, and the smallest number was about 0.6 millimeters.
For the commissioning part, we followed the AAPM Medical Physics guidelines, MPPG 5.a publication that has a number of tests that one should perform to commission the planning system. And what you see here is screenshots of three of the analyses were small, off-axis, and large fields. And the graphs are a little small, but at full size they again are small themselves. So for a small field size, we compared the measured dose with the calculated dose using the Octavius 1000, that's an SRS dedicated 2d dosimeter. And the spacing between the measurement points are 2.5 millimeter center to center.
So, we got a very good agreement for the small field. Same was true for the larger field, which was about an eight by eight field. And also for the off-axis field, that was about four centimeters off-axis as well. So, in conclusion of that part of the commissioning, we found the 2d gamma analysis had a better than 90% passing rates for all points that we evaluated with 2%/2 millimeter evaluation criteria.
We also got the IROC spine Phantom that we irradiated after developing the plan for Elements for this anatomy. And we delivered 6Gy with a maximum dose of less than 5.25 as prescribed by IROC to the cord. And this is here the results that we obtained back from IROC. Also, a high passing score that shows the match between the Elements calculation and the film, the film being the blue, and the purple being the Elements, respectively. So, we felt pretty confident that the commissioning had gone well. We continued by doing some patient-specific QA. At least one of the patients that we [inaudible 00:10:34] using the SRS 1000 Phantom. Analyzed with global gamma relative to maximum dose in volume, 3%/1 millimeter in this case, and passing rates were over 95%. As you can appreciate, qualitatively, at least from the graphs [inaudible 00:10:52] shown here for the patient using the slides.
We also used another Phantom, we used a Delta4. We did plan in Elements, and then we mapped the Delta4, we calculated, and did the analysis as well. And yet another Phantom, the one from RTsafe, this is a true replica of one of the patients that we treated that was made using a 3D printer. We filled the [inaudible 00:11:25] here with water. We had an ion chamber of A16 in the spine and inside the tumor. So, when we did the analysis for that, these are the results that we received for the Delta4 comparison. Also a very high passing rate of 3%/1 millimeter, over 95%. And at the bottom, you see the difference that we measure in the CTV in the spinal cord between the plan and the measured values, also in very good agreement. So, we've been pretty pleased with the agreement that we've been receiving.
So, in conclusion, the Brainlab Spine Elements is uniquely suited for spine SRS from our experience. It has several unique, novel features that take advantage of the spine anatomy and it's specific to the spine curvature correction, automatic target delineation, optimization for complex spine target shapes, and intuitive planning tools that greatly facilitate automation. And also Monte Carlo calculation was important, especially in the presence of bone and for regions close to lines where you have two levels of inhomogeneity to account for.
Spine Elements generated plans, produced statistically significant better cord sparing and target conformity when prepared in a planning study with the Monte Carlo of Monaco planning system and the Pinnacle. And for the commissioning, we conducted with a combination of beam model verification, according to [inaudible 00:13:06] guidelines, the IROC practice accreditation, Phantom, the RTsafe, and patient-specific measurements, and we found that we had very good results. And since then, we have treated several patients very successfully with the same type of QA as shown here, with high confidence that we deliver exactly what we plan. And with that, I'd like to thank you for your time.