Transcript
So last year at the symposium, I talked about our validation and commissioning efforts, and also patient-specific QA methods. This year, I'm here to present some example cases, clinical cases, and some challenging cases that we had at UCLA. These are my disclosures, and credits to my colleagues who either directly or indirectly contributed to this talk, many of them are in the audience today. So, if you noticed on the first slide, I'm also a quality officer of the UCLA Health. And so I represent our department in the UCLA Health System. This is one of the slides I used last year, we have a yearly report to the UCLA Health System, where we talked about a methodology that was implemented at our institution, that really improved patient experience. It also provided benefits, as well as increased throughput on the machine. Literally, if we had 11, 12, or 13 metastases that we had to treat for a patient, that basically meant long treatment planning process. That also means the physicist and a physician have to make 12 or 13 trips to the machine to do an iso check, image check. I've been on many of this treatment, and it's really fantastic to see a patient that would otherwise have been treated in two or three days, each day one or two hours, all of a sudden is done in 20 or 30 minutes. It's just fascinating. So, it has made a very big impact on our patients.
So, the first case that I'm presenting here is somewhat challenging in the sense of the quality assurance that we had to do. This was a patient with five metastases, three of them were really small, and the two of them were relatively larger targets. And this patient fell within the first 12 or so patient that we treated. The first 10 or 12 patients, we did very extensive QAs before we treated the patients. As you can see in the next two examples that I'll give you, amount of QA that we do per patient has dramatically decreased over time. So, this patient, we first transferred the plan into Eclipse. And we did a forward calculation in Eclipse to compare the dose distribution in Eclipse to multiple metastasis element. And what I'm showing you here, the five targets, and each of those lines that you see here is a reference point per target, compared with two planning systems showing clinically acceptable results. In the past, I've shown you that, in general, we expect within 5% agreement between two planning systems. In our case, our model in Eclipse is not tweaked for small targets. So, for smaller targets, which is the sort of the left of the graph, you see larger differences than for the large targets.
We also did a MU check calculation, some institutions use RadCalc. And with that, we also got pretty good results. Sun Nuclear has a very nice product, it's PerFRACTION. They have multiple ways that you can do a secondary check on your plan. One of them which we really like is the log file-based analysis, which means you deliver the plan, you don't even have to have a phantom in there, then you gather the log file and you back-calculate the dose distribution in a patient. Or you can do a secondary dose calculation and compare. So, I'm showing you the results for this patient were for all five targets, we were able to achieve less than 5% differences, and with a GammaCalc, 2%, 2 millimeters close to 100% passing rate. And then the two of the large targets, we decided to measure with an iron chamber. And as you can see, Mark was referring to this, the openings are not really that modulated. These are dynamic conformal arcs, and not all targets are covered because each pass, the system selects which targets will be covered. And the two targets that we decided to measure, which are these two targets I'm showing you here, the agreement was less than 3%. But then what do you do with the other three targets that are small? So with dose, what we did is we measured all five with an extended dose film using absolute values, not normalized. And for all five we were able to achieve close to 100% passing rate with a 2 millimeter and 2% criteria.
The second case I'm presenting was a challenging case where we had seven metastasis. One of them was in a brainstem, which our physician wanted to treat with a conical collimator to 12 gray, and I'm showing you that one here. The second one was pretty large one in cerebellum, where he wanted to treat that one with the 6 gray times 5, basically 30 gray in five fractions. So, I'm showing you the conical target here, and then the SBRT target on the top of here. It gets a bit more challenging because there was another target that was nearby the large target that we were considering to treat. And we were wondering whether we should go with a single fraction or fractionated scenario for that one. And at the end, we decided that that will be incorporated with the large target with the 6 gray times 5. And all of the other targets that I'm showing you here were decided to be treated with the multiple metastasis element. So, at the end, we treated the SBRT separately from MME. SBRT was five fractions. MME was one fraction, and the total was six fractions, which has its own consequences, right, because now you exceeded five fraction limit, and that may have some consequences.
Another approach you may have, which I'm showing you here is to cover the large target with an MME. So, what I'm showing you here is the large target along with a small one next to it, is covered in the multiple metastases element planning. So, you have all the other ones covered by 18 gray, and this large one 6 gray in the same plan. And then what you do is you utilize the cranial SRS element, which is still in elements to deliver four additional fractions of 6 gray to that target. So this has covered 6 gray in the initial plan with other targets, and then you deliver another four fractions, which makes it total of five fractions. And all of that is done with 18 elements. It's nice that way because the data transfer is seamless, and also there is no loss of information when it's resampled when you export and import into other planning systems. Also, in cranial SRS, you can consider the doses that were delivered to risk objects in the initial plan in a multiple metastases element that can be inputted into optimization. Remember, cranial SRS is a VMAT approach. This is a bit different than multiple metastasis element. Here, you're doing optimization for VMAT delivery.
And when you optimize with the VMAT in cranial SRS, the trajectories of arcs are also optimized. And you also have control over whether you want to shorten an arc to avoid certain metastasis. In this example, we made this arc shorter so that we avoid that one there. But because the table positions or arc trajectories are also optimized, they may coincide with the ones that you treated with the multiple metastases element, you can also change that. So, here, what I'm showing you here in a composite plan is that we forced the cranial SRS element to come up with a plan that has tabled positions that are different than MME. And at the end, you have this composite sort of delivery for MME plus cranial SRS. And in this case, you can deliver the whole thing in five fractions because one of the five fractions you covered in the multiple metastases element. And this is how the physical dose summation looks like. I understand that there are biological considerations here. One was five fractions, the other one was one. But in general, this is how it looks like when you sum the dose. Thirty gray here and 18 gray to the other ones. Yeah. And we did a QA for this one, there was no measurements on the machine, all we did is secondary calculation with the Eclipse and we compare to the original calculation. We also did the PerFRACTION calculation by Sun Nuclear. And you can see pretty good results here, you have less than 3% differences and about 100% passing rate for gamma calculation.
What I'm showing you here, so the impressive results that Mike was showing you. That patient that 6 gray times 5 fractions, this is the target, and this is the two-month follow-up image that I just obtained a couple of days ago before this symposium. So, that's the original plan and this is the two-month follow-up, and you can see the drastic change in the SBRT target. Also, look at one of the MME targets, this is the 18 gray target that you see here, and this is the two-month follow-up of that. So, this is pretty impressive stuff. And even the one in a brain stem in [inaudible 00:09:54] where we deliver to 12 gray with single faction using a 6 millimeter collimator, and you can see radiological response to that one as well. So a good testament of that targeting accuracy. Yeah. The third case on presenting which was a bit challenging is a seven metastasis case where one of the targets was well separated from the other targets. In this case, if you decide that you include that one target in the other cluster, what happens is the isocenter sort of goes away from the cluster, and it's away from all of the targets. It's somewhere in the middle here.
What that does is, all of the targets become farther away from the isocenter, so rotational errors may have bigger impact. Also, the targets may be covered with the thicker leaf pairs, because all of a sudden, now we were using HD-MLC if you're far away from the isocenter. It's not the 2.5-millimeter, but the 5-millimeter MLCs that cover that target. Also, for some of the arcs, that one target may actually fall outside of the radiation area. Remember, HD-MLC has a limitation of how far you can treat. The previous speaker already mentioned, when we have small targets that are nearby, and we include that into MME. For the V12 consideration, we consider the contiguous V12 volume of both of those together for our analysis. And in fact, we even include the GTV volume in that to make it a bit more conservative. I do need to mention that, although I had that slide comparing the single iso multiple metastasis treatment with the multiple isocenter treatment, the patient is not exactly getting the same treatment. When you do single isocenter treatment, we add margins.
So, first of all, the patient is getting a bit larger radiation volumes compared to multiple isocenter treatment. The other thing that happens, again, mentioned in one of the talks, the distribution that the patient gets with the multiple metastasis element is a bit more inhomogeneous. There is more hotspots inside the target compared to the other ones, which may be actually good thing for this patient. I'm just telling you the difference. So, here I'm showing you how if you include this target in the original set, you may have thicker leaves covering that. So, you can designate arcs for that one target, the new version of the software allows that. You cannot add arcs and say, "This arc will only cover that target." That's probably more significant feature for agility. MLC dynamic tracking is possible. Yeah. So, at the end, what we have decided to do is we decided to cover that one target separately with its own isocenter, with its own arcs. And that's how the distribution will look like. We have everything else together, and then that one is separate from the others. So, there are three scenarios here, if you include it in the initial sort of set of targets, then thicker leaves cover that. This is what conformity index and gradient index you get. The second option that you have is you basically include another isocenter routine MME to cover that target, then you have thinner leaves covering that. Or you can decide that that one target will be covered by cranial SRS different application. And in that case, it's a VMAT optimization. And you will have even better conformity index and gradient index. So, increasingly better from here to here. In fact, in this case, you may even decide not to use as big of margins, because now the isocenter is within the target.
So, at the end, that's how the arcs look like, and that's how the composite [inaudible 00:13:48] distribution looks like. Yeah. So I have shown this slide in the past where we do know that when you have isocenters per target, the rotational layers are not that significant. But when you have one isocenter for all of the targets then rotational layers become much more significant for those symmetric purposes. Half-degree rotation will cause a 0.5-millimeter misalignment, just because of that. So, we used 0.5-millimeter and 0.5-degree tolerances for our ExacTrac when we reposition this patient. So, in summary, we use 0.5 degrees, 0.5 millimeters, we add 1-millimeter or 2-millimeter margins, depending on how far away from the isocenter you are. Maybe Brainlab wants to implement that where you can specify targets that are within 5 cm, for example, we'll get 1-millimeter margin, the targets that are way more than 5 cm then you add more margin. And we can see there are more than one isocenter and as I showed you in our examples, for cases where it may be beneficial. At the end, all of these cases, we upload to cranial SRS registry. And we're encouraging everyone to do that so that we can have some collective analysis of our cases, and we'll learn more from our patient cases. Thank you for your attention.
So, the first case that I'm presenting here is somewhat challenging in the sense of the quality assurance that we had to do. This was a patient with five metastases, three of them were really small, and the two of them were relatively larger targets. And this patient fell within the first 12 or so patient that we treated. The first 10 or 12 patients, we did very extensive QAs before we treated the patients. As you can see in the next two examples that I'll give you, amount of QA that we do per patient has dramatically decreased over time. So, this patient, we first transferred the plan into Eclipse. And we did a forward calculation in Eclipse to compare the dose distribution in Eclipse to multiple metastasis element. And what I'm showing you here, the five targets, and each of those lines that you see here is a reference point per target, compared with two planning systems showing clinically acceptable results. In the past, I've shown you that, in general, we expect within 5% agreement between two planning systems. In our case, our model in Eclipse is not tweaked for small targets. So, for smaller targets, which is the sort of the left of the graph, you see larger differences than for the large targets.
We also did a MU check calculation, some institutions use RadCalc. And with that, we also got pretty good results. Sun Nuclear has a very nice product, it's PerFRACTION. They have multiple ways that you can do a secondary check on your plan. One of them which we really like is the log file-based analysis, which means you deliver the plan, you don't even have to have a phantom in there, then you gather the log file and you back-calculate the dose distribution in a patient. Or you can do a secondary dose calculation and compare. So, I'm showing you the results for this patient were for all five targets, we were able to achieve less than 5% differences, and with a GammaCalc, 2%, 2 millimeters close to 100% passing rate. And then the two of the large targets, we decided to measure with an iron chamber. And as you can see, Mark was referring to this, the openings are not really that modulated. These are dynamic conformal arcs, and not all targets are covered because each pass, the system selects which targets will be covered. And the two targets that we decided to measure, which are these two targets I'm showing you here, the agreement was less than 3%. But then what do you do with the other three targets that are small? So with dose, what we did is we measured all five with an extended dose film using absolute values, not normalized. And for all five we were able to achieve close to 100% passing rate with a 2 millimeter and 2% criteria.
The second case I'm presenting was a challenging case where we had seven metastasis. One of them was in a brainstem, which our physician wanted to treat with a conical collimator to 12 gray, and I'm showing you that one here. The second one was pretty large one in cerebellum, where he wanted to treat that one with the 6 gray times 5, basically 30 gray in five fractions. So, I'm showing you the conical target here, and then the SBRT target on the top of here. It gets a bit more challenging because there was another target that was nearby the large target that we were considering to treat. And we were wondering whether we should go with a single fraction or fractionated scenario for that one. And at the end, we decided that that will be incorporated with the large target with the 6 gray times 5. And all of the other targets that I'm showing you here were decided to be treated with the multiple metastasis element. So, at the end, we treated the SBRT separately from MME. SBRT was five fractions. MME was one fraction, and the total was six fractions, which has its own consequences, right, because now you exceeded five fraction limit, and that may have some consequences.
Another approach you may have, which I'm showing you here is to cover the large target with an MME. So, what I'm showing you here is the large target along with a small one next to it, is covered in the multiple metastases element planning. So, you have all the other ones covered by 18 gray, and this large one 6 gray in the same plan. And then what you do is you utilize the cranial SRS element, which is still in elements to deliver four additional fractions of 6 gray to that target. So this has covered 6 gray in the initial plan with other targets, and then you deliver another four fractions, which makes it total of five fractions. And all of that is done with 18 elements. It's nice that way because the data transfer is seamless, and also there is no loss of information when it's resampled when you export and import into other planning systems. Also, in cranial SRS, you can consider the doses that were delivered to risk objects in the initial plan in a multiple metastases element that can be inputted into optimization. Remember, cranial SRS is a VMAT approach. This is a bit different than multiple metastasis element. Here, you're doing optimization for VMAT delivery.
And when you optimize with the VMAT in cranial SRS, the trajectories of arcs are also optimized. And you also have control over whether you want to shorten an arc to avoid certain metastasis. In this example, we made this arc shorter so that we avoid that one there. But because the table positions or arc trajectories are also optimized, they may coincide with the ones that you treated with the multiple metastases element, you can also change that. So, here, what I'm showing you here in a composite plan is that we forced the cranial SRS element to come up with a plan that has tabled positions that are different than MME. And at the end, you have this composite sort of delivery for MME plus cranial SRS. And in this case, you can deliver the whole thing in five fractions because one of the five fractions you covered in the multiple metastases element. And this is how the physical dose summation looks like. I understand that there are biological considerations here. One was five fractions, the other one was one. But in general, this is how it looks like when you sum the dose. Thirty gray here and 18 gray to the other ones. Yeah. And we did a QA for this one, there was no measurements on the machine, all we did is secondary calculation with the Eclipse and we compare to the original calculation. We also did the PerFRACTION calculation by Sun Nuclear. And you can see pretty good results here, you have less than 3% differences and about 100% passing rate for gamma calculation.
What I'm showing you here, so the impressive results that Mike was showing you. That patient that 6 gray times 5 fractions, this is the target, and this is the two-month follow-up image that I just obtained a couple of days ago before this symposium. So, that's the original plan and this is the two-month follow-up, and you can see the drastic change in the SBRT target. Also, look at one of the MME targets, this is the 18 gray target that you see here, and this is the two-month follow-up of that. So, this is pretty impressive stuff. And even the one in a brain stem in [inaudible 00:09:54] where we deliver to 12 gray with single faction using a 6 millimeter collimator, and you can see radiological response to that one as well. So a good testament of that targeting accuracy. Yeah. The third case on presenting which was a bit challenging is a seven metastasis case where one of the targets was well separated from the other targets. In this case, if you decide that you include that one target in the other cluster, what happens is the isocenter sort of goes away from the cluster, and it's away from all of the targets. It's somewhere in the middle here.
What that does is, all of the targets become farther away from the isocenter, so rotational errors may have bigger impact. Also, the targets may be covered with the thicker leaf pairs, because all of a sudden, now we were using HD-MLC if you're far away from the isocenter. It's not the 2.5-millimeter, but the 5-millimeter MLCs that cover that target. Also, for some of the arcs, that one target may actually fall outside of the radiation area. Remember, HD-MLC has a limitation of how far you can treat. The previous speaker already mentioned, when we have small targets that are nearby, and we include that into MME. For the V12 consideration, we consider the contiguous V12 volume of both of those together for our analysis. And in fact, we even include the GTV volume in that to make it a bit more conservative. I do need to mention that, although I had that slide comparing the single iso multiple metastasis treatment with the multiple isocenter treatment, the patient is not exactly getting the same treatment. When you do single isocenter treatment, we add margins.
So, first of all, the patient is getting a bit larger radiation volumes compared to multiple isocenter treatment. The other thing that happens, again, mentioned in one of the talks, the distribution that the patient gets with the multiple metastasis element is a bit more inhomogeneous. There is more hotspots inside the target compared to the other ones, which may be actually good thing for this patient. I'm just telling you the difference. So, here I'm showing you how if you include this target in the original set, you may have thicker leaves covering that. So, you can designate arcs for that one target, the new version of the software allows that. You cannot add arcs and say, "This arc will only cover that target." That's probably more significant feature for agility. MLC dynamic tracking is possible. Yeah. So, at the end, what we have decided to do is we decided to cover that one target separately with its own isocenter, with its own arcs. And that's how the distribution will look like. We have everything else together, and then that one is separate from the others. So, there are three scenarios here, if you include it in the initial sort of set of targets, then thicker leaves cover that. This is what conformity index and gradient index you get. The second option that you have is you basically include another isocenter routine MME to cover that target, then you have thinner leaves covering that. Or you can decide that that one target will be covered by cranial SRS different application. And in that case, it's a VMAT optimization. And you will have even better conformity index and gradient index. So, increasingly better from here to here. In fact, in this case, you may even decide not to use as big of margins, because now the isocenter is within the target.
So, at the end, that's how the arcs look like, and that's how the composite [inaudible 00:13:48] distribution looks like. Yeah. So I have shown this slide in the past where we do know that when you have isocenters per target, the rotational layers are not that significant. But when you have one isocenter for all of the targets then rotational layers become much more significant for those symmetric purposes. Half-degree rotation will cause a 0.5-millimeter misalignment, just because of that. So, we used 0.5-millimeter and 0.5-degree tolerances for our ExacTrac when we reposition this patient. So, in summary, we use 0.5 degrees, 0.5 millimeters, we add 1-millimeter or 2-millimeter margins, depending on how far away from the isocenter you are. Maybe Brainlab wants to implement that where you can specify targets that are within 5 cm, for example, we'll get 1-millimeter margin, the targets that are way more than 5 cm then you add more margin. And we can see there are more than one isocenter and as I showed you in our examples, for cases where it may be beneficial. At the end, all of these cases, we upload to cranial SRS registry. And we're encouraging everyone to do that so that we can have some collective analysis of our cases, and we'll learn more from our patient cases. Thank you for your attention.