Transcript
So, we'll be switching some gears, and we'll be talking about patient-specific QA and IGRT requirements. Essentially, this is going to be two talks combined into one. So, disclosures and credits to many of my colleagues who either directly or indirectly contributed to this talk. First, I would like to share with you that the implementation of the multiple metastases element in our clinic has been very successful, with an improved patient experience, with a significant reduction in treatment times for our patients. It also provide benefits, where, let's say if you were treating 11 isocenters, the physician and the physicist had to make 11 trips to the machine in multiple days, actually, not in a single day.
Now you just go there once, and the treatment lasts only 15 or 20 minutes. It also increased the throughput, and I'm hearing that at some institutions, also the eligibility of the patients has increased with this treatment technique. This treatment technique and the software has been validated in many different ways. And I'm just showing you a quick summary of work that we have done with multiple vendors, and I will not be going into details of this. We have also summarized and worked with many vendors to come up with the patient-specific QAs for testing the plans per-patient basis. Again, you don't need to apply all of this, but hand-selected ones can be done.
I will be going through three different case scenarios. And as I go through those cases, you will see that the patient-specific QA that we have applied decreased over time, as our confidence in the software has increased. So the first case is a case with five metastases, where two of the targets were relatively large, and the three other ones were small. So, the larger ones, you can measure with an ion chamber. That's the reason why I'm presenting this case. So, we transferred the plan into another treatment planning system, and we forward calculated the dose distribution. And we compared the distribution between two planning systems. And there was a pretty good agreement, as you can see on the screen. There were some small discrepancies, but we knew from our previous research that as the target size decreases, going from this direction, going from large targets to smaller targets, there is some discrepancy between two planning systems. If you tweak the second planning system, you will get better agreements for smaller targets as well. Mu check was also applied for this plan similar to RadCalc, and the results were clinically acceptable.
We also tested Sun Nuclear secondary calculation algorithm for this, and the results were just outstanding as you can see here. The gamma analysis with 2% on 2-millimeter shown on this side here, we're almost 100% passing rate. So the two larger ones that we measured with the ion chamber just to validate the absolute dose, and as you can see on the screen, the agreement between the measured and the calculated dose was less than.... this disagreement was less than 3%. So then...By the way, I should mention that, as you can see from the MLC movements, these fields are not that modulated. It's not a volumetric modulator arc therapy. It's just a dynamic form arc. Yeah. So, now we have measured the two large targets and validated that. What do we do the other three? So we decided that we're going to measure all five of them with the film, and I'm showing you the results of IBT film measurement for all five targets. With great care, you can measure absolute dose with the film as well. So, as you can see here, there is a pretty good agreement for all five targets, including the small ones.
So the second case was a really challenging case, selected by our chairman for treatment. There were seven metastases. One was in a brainstem, and he required that we wanted to treat this with the conical collimator. And then there was a large one which he wanted to do an SBRT treatment, 6 gray times 5, so 30 in five fractions. And it was even more complicated because there was a small target right next to this SBRT target. And then there were another four that he wanted to treat with the multiple metastases element. So, things get complicated once you start actually using this software in clinic. So, what we decided to do is treat the four small targets to 18 gray. And then maybe you could include the 6 gray or 30 gray target, which is the SBRT target. In the initial plan of MME, multiple metastasis element, they only deliver 6 gray to that target. And then an additional...I should go here. Another additional four fractions of 6 gray can be delivered to that target. This is done strategically because this way, the total treatment for these targets will be five fractions, and at least in the United States there are billing really consequences if you go beyond five fractions. If you don't include the first fraction of 6 gray in that plan, then you will be ending up with a six fractions that you deliver.
So, as you can see here, that's the multiple metastases element plan that includes the 6 gray, and then another additional 4 times 6 gray with the other cranial SRS element. If you use cranial SRS element in conjunction with the MME, then you have a flexibility of trying to avoid similar arcs. So you can organize the arcs in a way that you don't pass through the same area twice. Also, the dose transfer between two elements is higher fidelity compared to, if you try to use another software by DICOM exporting and importing. So, this case, we also compared with a Eclipse treatment planning system, and the disagreements were less than 3%. We run through the dose check using Sun Nuclear dose check algorithm, and again, the results were clinically acceptable. So, then, I often get this question of, how do you QA the 4-millimeter collimator for that treatment of that metastasis there? If you do Winston-Lutz test with the 4-millimeter collimator, the ball itself is 5-millimeter, so this is a difficult thing to do. There are a few ways you can do this. One is you do the Winston-Lutz test with the 7.5-millimeter collimator, then you leave the 7.5-millimeter collimator there and you obtain an EPID image of that. And then you replace the cone with the 4-millimeter collimator, and then you acquire another EPID image of that, and you overlay them just like I did on the right-hand side here.
So the idea is you already QA the 7.5 cone, and if the agreement between the other two cones is less than 0.1, or actually when we measured it was less than 0.05 millimeters, then you just essentially QA the 4-millimeter collimator. So here's a sort of schematics of that showing that you measure with the 7.5-millimeter, then you measure with the 4-millimeter, and then you overlay it and analyze it for coincidence. So this is one way you can do it, or you can do it in a very simplistic way. You can put a film on the way of the beam, pin mark where the lasers are, you radiate the film, and then analyze the film. I'm showing you the worst-case scenario from that measurement on this side here. And as you can see, the pin mark is about 0.5, 0.6 millimeters away from the isocenter. So this is that same case, with a two-month follow-up. So this is the original plan and that's the follow-up image there, showing a significant reduction in the tumor size for the SBRT target. The MME target is almost completely resolved. And I also have an image of the metastases in the brainstem. So, there is a radiological response of that one as well. In vivo confirmation that we can target very small targets.
The third case that I'm showing you here has a cluster of metastasis inferior in a patient, and then there is one superior target, which is further away from all other metastases. If you include this metastasis in a multiple metastases element planning, then essentially what happens is that you increase the distances from isocenter to each of these targets. And what will happen is, the rotational misalignments will become dosimetrically more significant. Another thing that will happen is, as shown here, that remote metastasis will be covered by leaf pairs that are thicker than the leaf pairs in the center of the field. So you may be compromising the conformity and gradient index of the plan. So, also when we have two targets that are nearby each other, the way we analyze this is we look at the contiguous V12, not separately taken per target, but contiguous V12 to be less than 10 CC. We are actually a bit more conservative. We include the GTV volume in that analysis.
So, here's that plan where the inferior targets were planned with the multiple metastases element, and the superior target was planned separately. So, actually, you have several choices. The one that you see here is the target was included in the initial multiple metastases element planning. This choice here, what you do there is you assign another isocenter for that target, specifically, and you get better conformity because now you're using thinner leaf pairs there, and the conformity ingredients are better. Or you can plan that one target with the cranial SRS and get even better conformity index and gradient index. So you have all these choices.
Now, switching to IGRT requirements. In the past, I have communicated with you that when we have isocenters per target, rotational layers are not that detrimental for these cases. But when you have one isocenter, and you want to treat all these metastases, small rotational layers, this is the same rotation, as I showed you in the other case, are more detrimental. So, because of this, we have showed this graph where we said if the target is 6 centimeters away from the isocenter, and the rotation is 0.5 degrees, then that alone itself translates to 0.5-millimeter of discrepancy. But then the question was, well, this is theoretical analysis, how did the patients move? So, what we did is...Well, let me cover this first. Because of that analysis, we have decided that our criteria threshold for repositioning the patient will be 0.5 degrees and 0.5 millimeters. And that's what we use in a clinic. And we also use 2-millimeter margins to account for all of these uncertainties. And I will be showing you some preliminary data where we may want to reconsider that 2-millimeter margin. Maybe that's too large of a margin that we use.
So, here's an analysis of 74 targets, I think, that we analyzed. So, these patients required image prior to every arc that we treated. And we are looking at motion pattern of these patients. And if you look at the lateral, longitudinal, and vertical movements of these patients, the 75th percentile of this motion is less than 0.4 millimeters. Overall, if you include all the data, 98th percentile of the motion is 0.7 millimeters. And by the way, this motion was there, these translational uncertainties were there, even when we were treating with isocenter per target. So if you were not putting margins in the other case, there is no reason to add margin just because of this. This was there to start with. What's new is the rotations. So here, we are looking at that, again, the same 74 targets. Again, we're looking at pitch, roll, and yaw. And 75 percentile of these patients is less than, again, 0.4 degrees of rotation. Overall, 90th percentile, 0.6 degrees. So this is what needs to be considered for adding additional margin. So going back to analysis of, okay, how far are the targets, in general, from the isocenter in the clinical practice? This is the distribution that we came up with the actual patients that we treated.
So it looks like our initial assessment that between 4 centimeters and 6 centimeters away is where the targets are. That falls to here. So going back to this graph, if the target is 6 cm away, which is the 75th percentile of that bell curve that you saw, it's only additional 0.5 millimeters uncertainty that the rotations add. And in this slide, I just want to bring to your attention that going from zero-millimeter margin to 2-millimeter margin, we actually double if not triple the size of the treatment volume. We double if not triple the size of V5, V10, and V12. So it may seem like a small addition to the target, but, in general, in terms of volume, this is tripling the size of the target, tripling the size of the V5, V10, and V12. So, I guess with this preliminary data, I would like to suggest that maybe we want to reconsider the use of a 2-millimeter margin, especially considering that the multiple metastases element creates very inhomogeneous heart distributions. So now we're treating the patients with wider margins and higher doses. They're not the same treatments as what we were delivering in the past.
So, one more thing that we looked into is, with the 0.5-millimeter and 0.5-degree threshold, we looked into to see how often do the therapist reposition the patient. And it turns out with that criteria, the patients are repositioned about 50% of the time. I know the software doesn't allow it to go any smaller than that, but if you were to go down in this direction, you would basically chasing a noise. You would be repositioning the patient all the time. And going the other way around, you would allow large patient motions, which not be acceptable. So, we still stand that 0.5 and 0.5 is the probably most clinically acceptable thresholds to use. And by the way, this was in agreement to our previous data from trigeminal patients. In this case, we had a larger cohort, but again, with a 0.5 threshold, but 50% of the time that therapist is repositioning the patient. So I guess the thresholds that we use, it seems to me are clinically relevant and correct, but the margins that we use may need to be reconsidered. That's all. Thanks.
Now you just go there once, and the treatment lasts only 15 or 20 minutes. It also increased the throughput, and I'm hearing that at some institutions, also the eligibility of the patients has increased with this treatment technique. This treatment technique and the software has been validated in many different ways. And I'm just showing you a quick summary of work that we have done with multiple vendors, and I will not be going into details of this. We have also summarized and worked with many vendors to come up with the patient-specific QAs for testing the plans per-patient basis. Again, you don't need to apply all of this, but hand-selected ones can be done.
I will be going through three different case scenarios. And as I go through those cases, you will see that the patient-specific QA that we have applied decreased over time, as our confidence in the software has increased. So the first case is a case with five metastases, where two of the targets were relatively large, and the three other ones were small. So, the larger ones, you can measure with an ion chamber. That's the reason why I'm presenting this case. So, we transferred the plan into another treatment planning system, and we forward calculated the dose distribution. And we compared the distribution between two planning systems. And there was a pretty good agreement, as you can see on the screen. There were some small discrepancies, but we knew from our previous research that as the target size decreases, going from this direction, going from large targets to smaller targets, there is some discrepancy between two planning systems. If you tweak the second planning system, you will get better agreements for smaller targets as well. Mu check was also applied for this plan similar to RadCalc, and the results were clinically acceptable.
We also tested Sun Nuclear secondary calculation algorithm for this, and the results were just outstanding as you can see here. The gamma analysis with 2% on 2-millimeter shown on this side here, we're almost 100% passing rate. So the two larger ones that we measured with the ion chamber just to validate the absolute dose, and as you can see on the screen, the agreement between the measured and the calculated dose was less than.... this disagreement was less than 3%. So then...By the way, I should mention that, as you can see from the MLC movements, these fields are not that modulated. It's not a volumetric modulator arc therapy. It's just a dynamic form arc. Yeah. So, now we have measured the two large targets and validated that. What do we do the other three? So we decided that we're going to measure all five of them with the film, and I'm showing you the results of IBT film measurement for all five targets. With great care, you can measure absolute dose with the film as well. So, as you can see here, there is a pretty good agreement for all five targets, including the small ones.
So the second case was a really challenging case, selected by our chairman for treatment. There were seven metastases. One was in a brainstem, and he required that we wanted to treat this with the conical collimator. And then there was a large one which he wanted to do an SBRT treatment, 6 gray times 5, so 30 in five fractions. And it was even more complicated because there was a small target right next to this SBRT target. And then there were another four that he wanted to treat with the multiple metastases element. So, things get complicated once you start actually using this software in clinic. So, what we decided to do is treat the four small targets to 18 gray. And then maybe you could include the 6 gray or 30 gray target, which is the SBRT target. In the initial plan of MME, multiple metastasis element, they only deliver 6 gray to that target. And then an additional...I should go here. Another additional four fractions of 6 gray can be delivered to that target. This is done strategically because this way, the total treatment for these targets will be five fractions, and at least in the United States there are billing really consequences if you go beyond five fractions. If you don't include the first fraction of 6 gray in that plan, then you will be ending up with a six fractions that you deliver.
So, as you can see here, that's the multiple metastases element plan that includes the 6 gray, and then another additional 4 times 6 gray with the other cranial SRS element. If you use cranial SRS element in conjunction with the MME, then you have a flexibility of trying to avoid similar arcs. So you can organize the arcs in a way that you don't pass through the same area twice. Also, the dose transfer between two elements is higher fidelity compared to, if you try to use another software by DICOM exporting and importing. So, this case, we also compared with a Eclipse treatment planning system, and the disagreements were less than 3%. We run through the dose check using Sun Nuclear dose check algorithm, and again, the results were clinically acceptable. So, then, I often get this question of, how do you QA the 4-millimeter collimator for that treatment of that metastasis there? If you do Winston-Lutz test with the 4-millimeter collimator, the ball itself is 5-millimeter, so this is a difficult thing to do. There are a few ways you can do this. One is you do the Winston-Lutz test with the 7.5-millimeter collimator, then you leave the 7.5-millimeter collimator there and you obtain an EPID image of that. And then you replace the cone with the 4-millimeter collimator, and then you acquire another EPID image of that, and you overlay them just like I did on the right-hand side here.
So the idea is you already QA the 7.5 cone, and if the agreement between the other two cones is less than 0.1, or actually when we measured it was less than 0.05 millimeters, then you just essentially QA the 4-millimeter collimator. So here's a sort of schematics of that showing that you measure with the 7.5-millimeter, then you measure with the 4-millimeter, and then you overlay it and analyze it for coincidence. So this is one way you can do it, or you can do it in a very simplistic way. You can put a film on the way of the beam, pin mark where the lasers are, you radiate the film, and then analyze the film. I'm showing you the worst-case scenario from that measurement on this side here. And as you can see, the pin mark is about 0.5, 0.6 millimeters away from the isocenter. So this is that same case, with a two-month follow-up. So this is the original plan and that's the follow-up image there, showing a significant reduction in the tumor size for the SBRT target. The MME target is almost completely resolved. And I also have an image of the metastases in the brainstem. So, there is a radiological response of that one as well. In vivo confirmation that we can target very small targets.
The third case that I'm showing you here has a cluster of metastasis inferior in a patient, and then there is one superior target, which is further away from all other metastases. If you include this metastasis in a multiple metastases element planning, then essentially what happens is that you increase the distances from isocenter to each of these targets. And what will happen is, the rotational misalignments will become dosimetrically more significant. Another thing that will happen is, as shown here, that remote metastasis will be covered by leaf pairs that are thicker than the leaf pairs in the center of the field. So you may be compromising the conformity and gradient index of the plan. So, also when we have two targets that are nearby each other, the way we analyze this is we look at the contiguous V12, not separately taken per target, but contiguous V12 to be less than 10 CC. We are actually a bit more conservative. We include the GTV volume in that analysis.
So, here's that plan where the inferior targets were planned with the multiple metastases element, and the superior target was planned separately. So, actually, you have several choices. The one that you see here is the target was included in the initial multiple metastases element planning. This choice here, what you do there is you assign another isocenter for that target, specifically, and you get better conformity because now you're using thinner leaf pairs there, and the conformity ingredients are better. Or you can plan that one target with the cranial SRS and get even better conformity index and gradient index. So you have all these choices.
Now, switching to IGRT requirements. In the past, I have communicated with you that when we have isocenters per target, rotational layers are not that detrimental for these cases. But when you have one isocenter, and you want to treat all these metastases, small rotational layers, this is the same rotation, as I showed you in the other case, are more detrimental. So, because of this, we have showed this graph where we said if the target is 6 centimeters away from the isocenter, and the rotation is 0.5 degrees, then that alone itself translates to 0.5-millimeter of discrepancy. But then the question was, well, this is theoretical analysis, how did the patients move? So, what we did is...Well, let me cover this first. Because of that analysis, we have decided that our criteria threshold for repositioning the patient will be 0.5 degrees and 0.5 millimeters. And that's what we use in a clinic. And we also use 2-millimeter margins to account for all of these uncertainties. And I will be showing you some preliminary data where we may want to reconsider that 2-millimeter margin. Maybe that's too large of a margin that we use.
So, here's an analysis of 74 targets, I think, that we analyzed. So, these patients required image prior to every arc that we treated. And we are looking at motion pattern of these patients. And if you look at the lateral, longitudinal, and vertical movements of these patients, the 75th percentile of this motion is less than 0.4 millimeters. Overall, if you include all the data, 98th percentile of the motion is 0.7 millimeters. And by the way, this motion was there, these translational uncertainties were there, even when we were treating with isocenter per target. So if you were not putting margins in the other case, there is no reason to add margin just because of this. This was there to start with. What's new is the rotations. So here, we are looking at that, again, the same 74 targets. Again, we're looking at pitch, roll, and yaw. And 75 percentile of these patients is less than, again, 0.4 degrees of rotation. Overall, 90th percentile, 0.6 degrees. So this is what needs to be considered for adding additional margin. So going back to analysis of, okay, how far are the targets, in general, from the isocenter in the clinical practice? This is the distribution that we came up with the actual patients that we treated.
So it looks like our initial assessment that between 4 centimeters and 6 centimeters away is where the targets are. That falls to here. So going back to this graph, if the target is 6 cm away, which is the 75th percentile of that bell curve that you saw, it's only additional 0.5 millimeters uncertainty that the rotations add. And in this slide, I just want to bring to your attention that going from zero-millimeter margin to 2-millimeter margin, we actually double if not triple the size of the treatment volume. We double if not triple the size of V5, V10, and V12. So it may seem like a small addition to the target, but, in general, in terms of volume, this is tripling the size of the target, tripling the size of the V5, V10, and V12. So, I guess with this preliminary data, I would like to suggest that maybe we want to reconsider the use of a 2-millimeter margin, especially considering that the multiple metastases element creates very inhomogeneous heart distributions. So now we're treating the patients with wider margins and higher doses. They're not the same treatments as what we were delivering in the past.
So, one more thing that we looked into is, with the 0.5-millimeter and 0.5-degree threshold, we looked into to see how often do the therapist reposition the patient. And it turns out with that criteria, the patients are repositioned about 50% of the time. I know the software doesn't allow it to go any smaller than that, but if you were to go down in this direction, you would basically chasing a noise. You would be repositioning the patient all the time. And going the other way around, you would allow large patient motions, which not be acceptable. So, we still stand that 0.5 and 0.5 is the probably most clinically acceptable thresholds to use. And by the way, this was in agreement to our previous data from trigeminal patients. In this case, we had a larger cohort, but again, with a 0.5 threshold, but 50% of the time that therapist is repositioning the patient. So I guess the thresholds that we use, it seems to me are clinically relevant and correct, but the margins that we use may need to be reconsidered. That's all. Thanks.