Transcript
Bocton:Today I have the privilege to introduce to you Manuel Todorovich, who is the chief of clinical medical physics at the Universitetsklinikum Hamburg Eppendorf, a hospital here. It is a Novalis certified institution. Mr. Todorovich is a very early adopter of our elements multiple brain mets SRS technology, as well as a longstanding Novalis customer since 2010. Today he will present to you his experience with our new multiple brain metastases elements, and discuss some of his clinical results.
Manuel: Thank you, Bocton, for this nice introduction. I'm happy to share my, or our experience, the European experience or more the Hamburg experience with the elements multiple brain mets SRS.
Bocton: Thank you.
Manuel: What are we going to talk about within the next couple of minutes, next hour, roughly. First of all, I would like to give you a short introduction about our department a little bit, so that you have an idea of what our department looks like, what kind of patients we do have and so on. And then a little outlook on the history of the multiple brain metastases element in our department. At which point did we get involved into development maybe and so on, and just to give you an idea why I want to talk about that. And then my talk is more or less split into three different parts. So the first one would be that I wanted to talk about what is needed to get theoretically a good dose distribution.
How can treatment planning be done more or less? In that terms, I would like to talk about what kind of patients are suitable, what kind of different options are there for these patients. We will have a quick, roughly look at two papers that are out there that might help us in decision-making. And then I would like to compare some kind of different treatment options that are out there for patients with multiple brain metastases. I just wanted to get you through the whole process that we did in our department when we got in touch with the software, and we thought about what is the purpose of the software, is it good, how can we deal with this and so on. If we can get theoretically a good dose distribution, then I would like to talk about how can we realize this dose distribution for the patient.
And then we have to talk about uncertainties little bit about image guidance. This is a very, very important part because as you know, we deal with one virtual ISO center, and at that point, we need to be sure that image guidance is working quite nicely and that we really hit what we want to hit. If we have done all the image guidance, then it's up to patient-specific QA. So how to get the dose really to the patient, and does the machine work properly and everything like that. I will tell a bit about the way we were doing that and how we can do it with the element solution. In the last part, I would like to give you some case reports of patients that we treated in our department here, and to give you an idea how everything would look like. Three or four patients I would like to present, and then a little bit some clinical results and a short outlook. That should be everything I want to present to you.
First of all, some key facts about our department. Down here is our Novalis STx system that is used for treating all these patients with brain tumors and brain metastases. We do have approximately 2000 new patients per year, and roughly 200 of them are really brain metastases patients. And we have another 100 brain tumor patients. These two indications are within the top five, that are treated in our department. So that's why we decided to use brain lab software, because we really, really believe in the technology. We are a Brainlab customer since 2000. As I said before, we have a Novalis STx system with the ExacTrac, the 60 robotics couch that is used to treat these patients. What about Multi Brain Metastases SRS at the UKE at our university medical center.
So, we first came in touch with Brainlab in the end of 2013. So roughly four years ago. They were here, and we had a usability evaluation of the really first prototype of the software. So they came up with the idea of treating multiple brain metastases with a virtual ISO center, and they asked us, what do you think about that? We tried it, we had some nice ideas, nice thoughts, and then we figured out how to get a research and development program together with Brainlab. In April, 2014, and then the first prototype was installed in our department. We were really doing a lot of development and research with that. And then another prototype was installed in May, 2015...on April, 2015. We treated our first patient using this prototype. The beginning 2017, we were the first European installation of the released clinical system. Right now, we are using the Cranial SRS, the Brainlab Multi Brain Metastases SRS, and also the Spine SRS software were installed in our department, and are clinically used.
We have more than 100 patients treated using the elements and the Brain Metastases SRS and over 20 patients treated using Cranial SRS, and numbers are increasing constantly. This is all about our department, so who should be treated. I think all of you know out there that in the past whole-brain radiotherapy was more or less the option that Linac-based clinics would offer patients with multiple brain metastases. Multiple brain metastases in the past meant most of the time, more than four or more than five lesions. So it was more like you got a diagnostic department, and they were counting the lesions and it was more like one, two, three, four, five, multiple. And whenever you got the diagnosis of multiple brain metastases, you used whole-brain radiation, that was in the past days.
But today I think, we all want to do it in a more local way by using SRS just because of preserving neurocognitive function and so on. Unfortunately in the past, the possibilities for this treatment were very limited. I mean, there was Gamma Knife out there for decades that were doing this as well, but there was no nice Linac-based approach for this. Just to give you an idea that brain mets is still still a hot topic. When I prepared this talk, I just did a little research on Google Scholar, and I just entered multiple brain metastases one isocenter, you can see it here. I just wanted to know how much papers roughly were published within 2016 and 2017 now. It came up with roughly 820 papers.
Brain metastases using one single-isocenter is really a hot topic, and it's really of interest out there, just to give you an idea that we are not talking about something that might be already not on vocal or anything like that. So what were the three major treatment options? I mean, among others, there are systemic treatments, chemo, and so on. But what are the major three major options? In the past, it was whole-brain radiotherapy, maybe with or without hippocampal sparing. We've got neurosurgery of single individual mets, then maybe combined with whole-brain radiotherapy or nowadays combined with SRS. I will give you a case within the case reports of that. SRS or STx, so hypofractionated therapy or single-fraction for multiple brain metastases alone. This could be combined with whole-brain radiotherapy if you want or not, but just to give something upfront, whole-brain radiotherapy will not help you at all. So there's no reason to implement that at all.
So what we thought about in our department is that everyone knows that whole-brain radiotherapy can be a toxic treatment. And the question is, do the benefits justify the risk in this situation? Therefore it was quite nice that Brainlab came to us with this nice idea of software. We, from day one on, were really interested and see how it works and how it could be used. What we did then, but in the past, we did a little study on papers out there just to have an idea of what we should do and how we could deal with this, and how to decide, which would be the best solution to deal with this disease for this patient to help a patient the best.
Whenever you talk about that, about multiple brain metastases, I think the first study that will come up will be the Yamamoto study. I will talk about that in the next minutes. Published already in 2014, there was an update just some months ago, but still the results are the same. Then there's a Kocher study in Germany already 2010, and there are a couple of other studies, but just focus first on the Yamamoto paper. This is a really nice paper because it's a multi-institutional study, and it really has roughly 1,200 patients in there. It's a really, really large study. All of these patients were treated with Gamma Knife. So this is a Gamma Knife study. They divided their patients into three groups. So one group was patients with only one single met, and the second group was patients with 2 to 4 mets, and the third group was patients with 5 to 10 mets, so something, what we would call multiple brain metastases in the past.
The results were that with 5 to 10 brain metastases, the treatment is non-inferior or the outcome for the patient is non-inferior to that with patients with two to four brain metastases. Two to four or two to three would be something that we already treated with a Linac-based approach in the past. This is something that we already knew that we can do good results on that. So very promising. It doesn't make sense to split groups there. So we can treat patients with up to 10 brain metastases, and our Gamma Knife community is still, already does that for 25 or so lesions in a row if they want to.
Just some key facts out of the study. The first one is that there were no grade 5, and almost no grade 4 adverse events during this study. Doesn't matter if it was 1 tumor, 2 to 4 or 5 to 10 tumors, all goes with the same. The next very important point is that the maintain...Kocher function was also the same for all of these groups. It was really, really nice. And you see, we got here numbers of 94% and 95% already for four months after SRS. The neurocognitive function was very nicely preserved for all of these patients. The last, maybe most important information is that if you look at the overall survival, you will see here, the blue and the red curve are for the 2 to 4, and 5 to 10 mets groups, and you see they are the same. So there's no change in overall survival for the patients. Doesn't matter if they got 2 lesions or got 10 lesions, it doesn't matter. They are all behaving the same. They have the overall survival curve, and there is no difference.
The one tumor group is a complete other story. So they are doing better, but this is more or less complete different disease here, because they are doing better in all other points of perspectives as well. So then just a quick look at another study, because I think this is a really nice paper because the idea behind this paper is does the number of tumors matter? It's a little bit like, do we really need to talk about number of lesions or do we have to talk about something else? They found out that the number is a really nice surrogate for the disease, but it's not an independent predictive factor. It's more like a surrogate for a lot of other clinical factors. So people with more lesions are doing worse just because of other diseases and other problems that they have as well. And then they came up with the idea that for patients with more than four brain metastases, SRS could be a good approach. It should be the future approach at all.
All these things convinced us already that we should go this way and follow the Brainlab road, and get more into detail with this. What we then did at that point is we used the prototype of the software and compared it to three other possibilities to do treatment planning and treatment for these patients, just because we wanted to know how good the software really is in comparison to the other things. We compared it to iPlan. So this would be an approach, multiple isocenter dynamic conformal arc, like we did in the past. For one, two, three lesions, we use single isocenter for every lesion, then use a couple of RX, or a couple of beams, and try to figure out, and try to arrange a good dose distribution at that point. We just try to do it for patients with up to 10 mets. So we use this on 16 different datasets.
We use Eclipse volumated arc therapy at that point, so RapidArc approach. We have another institution here in Hamburg that had had a Gammaknife, and that were offering us the possibility to do treatment planning for all these patients using the Gammaknife, because this was more or less the gold standard in the past, because they were doing this already for 20 years or so. So how did we do it? We did pre-contour all the datasets in Eclipse using...in elements using the automatic fusion and using all the nice algorithms and tools for outlining and delineation of the targets. And then we entered this pre-contour datasets into the three other treatment planning modalities, and in elements, we used one isocenter. So the virtual isocenter approach that will be positioned in the center of mass of all lesions. We used a template with up to 5 table angles, and that came up to 10 arcs in total, and this wasn't completely automatic approach. So just hit the calculate button and the software came up with a solution within three to five minutes.
For iPlan, as I said, the multi-isocenter dynamic conformal arc approach, we used one isocenter per lesion, three table angles with up to five arcs in total. This is a complete manual approach. So you have to do it all-forward planning. So start with the first lesion, arrange everything, at the next lesion, arrange everything, and just have to fool around a lot with table angles and so on to be sure that you do not influence each other too much, because that has to be taken into account as well. For the Eclipse approach, we used the same isocenter that the element came up with, we used 5 table angles as well, and also 10 arcs as well in total. So we used exactly the same setup, like brain metastases already came up, like the element already came up with. The only thing that we changed was we changed the column data field opening to allow the Eclipse software to cover all lesions within one arc if it thought it would be needful. So just to get more degrees of freedom for the software to be more convenient there.
For the Gammaknife, they used up to 10 shots. Was done with Leksell GammaPlan, up to 10 shots per lesion with maximum of 25 minutes beam on time per lesion. Let's have a look at the results. The next slides will all look the same more or less. So we will have the brain metastases on the left side, the dose distribution for that. The other solution on the other side that we would like to compare with, we got a little tabular down here to have an idea about the V12Gy. So something that might be V12 or V10Gy, that might be a good predictor for radionecrosis or so. So the volume of the brain receiving 12Gy, and we got the volume of the brain receiving 3Gy to get an idea about low-dose shower, or low-dose spillage at all.
This is just for one of the cases, to give an example. So it was six metastases cases, 1.5 CC tumor volume in total, and we prescribed 20Gy to the isodose covering the lesion. What you can see on the first look is that the brain metastases case is really nicely conformal. The blue area, the dose cut-off is at 3Gy in all cases. You see that the low dose shower is a lot lower than for the VMAT approach in this case. Even the V12Gy is half the amount of that, what came up to the VMAT approach. What you can see, it's nicely conformal. You can see already on the VMAT approach, that there will be another lesion in both situations here.
Another lesion, just a little bit ahead...more cranial to the cranial direction that the brain metastases solution can really spare and can really split the dose distribution to both. You already see that you get a high dose area here for the VMAT approach, so it's not the conformity at all compared to the other one. Same, if you look at the other dimension, you see the large, low dose shower here. You also see some three gray areas over here where there's not even in lesion in this case, in this situation. If we move over to the multiple isocenter dynamic conformal arc approach, you'll see that the dose distribution is also nicely conformal. You see no high dose areas here next to it, maybe a little bit here, but not here, next to it. The V12Gy is in the same magnitude and even the V3Gy.
So the low dose is only 200 CC compared to 700 something that we have for the VMAT. Even the low dose shower is a little bit lower. This also looks like a nice approach for doing treatment planning despite the fact that it's really time-consuming because you have to do it all-forward planning and have to spend a lot of time on that, and you will also need to spend a lot of time on the machine, but I will come to that in a couple of seconds here within the next slides. So good approach. Next comparison would be the Gammaknife. Sorry for having not the same visualization for the Gammaknife, but this is just due to the fact that the GammaPlan software is not able to export the dose distribution in a good way so that you can import it into any other software, Eclipse, for instance, what I used here.
In this case, I only pointed out here the 3Gy isodose line to give you an idea about the low-dose shower. We were only able to say that V12Gy and 3Gy are in the same magnitude for the GammaPlan and for the Gammaknife situation. I didn't get even numbers here. So because of the fact that the Gammaplan looks at separate dose distribution or dose cubes around each lesion and does not really offer a complete dose distribution or dose volume for the whole brain that is calculated really completely in detail. So we just end up with the same magnitude. To put that all together, I created a little comparison chart, or we created a little comparison chart in our department. We looked in this comparison at the conformity index. So this is inverse Paddick. It should be one to be perfect, same for the gradient index. We looked at the dose spillage in terms of the volume received in 3Gy.
We looked at the planning time and even at the delivery time and monitor units where applicable. Up here are the brain metastases numbers. You see, you can get a nice plan within 5 to 10 minutes, depending on if you have to arrange something a little...change an arrangement a little bit and hit the calculate button again. So 5 times 5 to 10 minutes was what the prototype took right now with a clinical system, we are more in the magnitude of 5 minutes or so for our plan. And it's even calculating a little bit faster. Delivery time is really interesting. As we only deal with one isocenter, all lesions were treated with the same isocenter.
The delivery time is only dependent on the prescription. So if you prescribe 23Gy, it takes a little bit longer than prescribing 18Gy or so. But you only have to apply that once. This is the biggest drawback for the multiple isocenter dynamic conformal arc approach. Because at that point, if you want to treat 10 lesions, you have to apply 10 times 18 or 23Gy. So you end up with this time multiplied by the number of lesions, and this will easily end up for 10 lesions with five hours or so, and this will really then kill all your schedule, and you will only be able to treat on patients. So that's one of the points and one of the reasons why in the past, no one using linear accelerator thought about treating multiple lesions within one session. There was no real solution for that.
The VMAT was...delivery time was the same, conformity index was also in the same magnitude, but the gradient index, so the dose fall off was worse, and the dose spillage as we have seen on the images was also very bad compared to the brain metastases case in my situation here, at least. Planning time was even longer because it was not an automatic process. Just calculating dose for dose grid of one millimeter just roughly takes 20 minutes or so. So just calculating is already two times the planning time for the whole-brain mestatases approach. Gammaknife, you see it's the same dose spillage might be the same, even better...I cannot really say something about that because the numbers were not in a good way prepared at that point. What I can say is the delivery time is worse because for 10 lesions, it takes you roughly two and a half up to three hours treatment time for the patients or beam on time on the table. The good point is the machine doesn't make any sound. So the patient can fall asleep during this time, but still the machine is occupied by the patient for time.
If we can do and provide good dose distributions theoretically, so we can use elements to get a nicely conformal dose plan for the patient. And we have to think about what comes if we move to the machine. We got a lot of non-coplanar RX. In the past, when we were dealing with non-coplanar RX, and we used the multiple isocenter approach, we had a lesion directly in the isocenter. Whenever we got a rotational error, and this can easily happen because the couch rotation is the weakest part in the whole chain and the whole treatment chain. When we did this in the past, that wasn't a very big deal because the lesion was in the isocenter, and if we introduce an error rotating around the isocenter, the resulting error was not that high.
Maybe we ended up adding a little margin as well. All the lesions were round-shaped. So that was not a real big deal or real big problem. But nowadays, if you use something like virtual isocenter, if you look at these, you see that there's no lesion in the isocenter at all. These are all the lesions in this case, the yellow, the orange, and the blue one, you see the leaf of field openings here were treated for this arc. And then just to get an idea of what happens, I introduced a little rotational error. I fooled around a little bit with the DICOM data and introduced a rotational error. What you can see is that the yellow, the orange and the blue one, they all move out of the field opening. You will miss these lesions. To come up with some numbers, I just did a calculation then of the plan.
I introduced translational errors here. You can see that the mean dose decreases by 2.5Gy in this case, and the minimum doses decreases almost by 7Gy if you got a 2-millimeter translation error and it already starts for one millimeter to really create problems. The same for rotational error. Rotational errors can really easily be introduced. And rotational error of one degree in pitch and roll, if you do know image correction or something like that or imaging, this will happen and you will not see it at all. You will end up with an under dosage here. Not even an under dosage for the lesion as well, you will end up with an overdosage to the normal brain because you will not hit the lesion, but the normal brain. It's getting worse if the distance of the lesion to the isocenter to the virtual isocenter increases, because it's just a geometrical problem. Just mathematics behind that that end up with this.
We need to have something to deal with this. The good point is we have nowadays frameless radiosurgery. There's a nice paper out there since 2011 by the Brussels Group by Thierry Gevaert who compared the frame-based approach, so with an head ring to the frameless approach. They ended up that frameless is even better than frame-based, just because of the fact that you introduce imaging. For the frame-based approach, you have no imaging. You rely completely on the fact that the patient is completely immobilized. If you move over to frameless, you have a mask, but you do imaging, do correction for that. At that point, you get even better than the frame-based approach. In their publication, they ended up with an error of roughly 0.7 millimeters in the 3D vector for all three orientation.
Frameless is a good approach and we can use it. How do we do it? We do it with ExacTrac. What I really recommend is whenever you change the table angle, and move over to a non-coplanar arc table angle, you should do imaging using ExacTrac. You should apply the corrections, and then you should do another set of images just to be sure that the correction was applied correctly, and then start treating that table angle again. They ended up with this 0.7 millimeters, and this is also just mathematics. This more or less depends on the imaging that you use for treatment planning. On the resolution of the CT, because it's based on the TRR images that are used in the ExacTrac. If you got a better resolution for the CT, for image planning, for treatment planning, you will end up with a bad result for the positioning because your TRRs will not be exactly in the Exac. At least you should use images of one-millimeter resolution, at least.
What can we do for quality assurance for this? We all know we have the Winston-Lutz test, which is used for the lasers mounted to find if they really are hitting the radiation isocenter defined by the linear accelerator, but now we introduce another system, the ExacTrac system, and this has its own imaging isocenter. This is created by the x-ray imaging and the infrared marker system. If you want to be exact and accurate, you need to be aware that your x-ray isocenter is exactly the same like your actual radiation isocenter of the accelerator. A good tool to check that is maybe a hidden target test or an end-to-end test. Just to give you a short idea, how we do it in our department, we use the STEEV Phantom here. We have a little BB in there, you can see it here. So from some of the CT slices, and we do the imaging, then we do the planning using the element. Like you can see here, we do the localization using the ExacTrac. We do the treatment delivery. At that point, we have an iron chamber or films inside of our Phantom. You do an analysis of that. In the end, we want to end up with one-millimeter end-to-end localization, and 2% dosimetric evaluation. This is something that we can achieve, and this is something that we do at least on a regular basis. But yeah, it can be done in this way.
Another point that I really would want to bring up is this really nice publication that I really would recommend to all of you out there. It's done by Tominaga Hirofumi, published already in 2014. They had an ExacTrac system and then Novalis STx, and they wanted to figure out if the ExacTrac system is really able for compensate for all these errors. This is something that we are also interested in. If we want to deal with this, we need a system that really is able to compensate for the rotational errors. Their approach was quite nice and quite easy. This is really beautiful. They created the little cubic Phantom with a couple of spheres in it. One was the target in the isocenter. This was the reference target. They introduced three other spheres here around this, these blue ones here. They were used for image-guided marker fusion, so that you could use the ExacTrac and also for Cone Beam CT. And then they just brought in some targets.
Four targets at 52 millimeters distance from the isocenter, these would be covered by the 2.5-millimeter leaves of the STx system, and another four targets at 87 millimeters distance, this would be covered by the 5-millimeter leaves in this case, so they could use the Phantom forces. Other things, I just want to look at the idea of if the patient positioning or the positioning accuracy can really be achieved by using this. What they did then is they positioned their Phantom correctly, did a Cone Beam CT, did ExacTrac, used the Cone Beam CT then as the gold standard, and then introduced errors. First of all, they introduced 0.5-degree error in roll and pitch. The Cone Beam CT at that point marker fusion, and ended up with a deviation of already 1 millimeter in the anterior-posterior direction. This one here.
And then they said, okay, but we can have 1 degree in roll and pitch as well. So 0.5 would be perfect, but 1 degree is really easily possible and can happen. They introduced that, did again a Cone Beam CT, and they ended already up with 2.3 millimeter in the anterior-posterior direction. If you look... Remember my DVH images that I showed you, you will see that this will already end up with an unbelievable under dosage to your lesion, and you will really miss what you wanted to hit at that point. What they then did is they did ExacTrac, did automatic fusion, did correction, rearranged the table position, did another ExacTrac, and did a Cone Beam CT, and then at that point, just a Cone Beam CT, just to see if everything is really in place, because that was the gold standard at that point, and they ended up with a maximum deviation of 0.6 millimeters.
ExacTrac was really able to compensate for all these things, put everything back to 0.6 millimeters, which is absolutely perfect within stereotactic dimension so everything should perfectly work, and is a nice tool. This is why you really should use this, and I really recommend to do imaging whenever you change the table here. So now we can theoretically do doses distribution by using elements. We have ExacTrac onboard, so we are pretty sure that we really hit what we want to hit. The last point in the whole chain would be patient-specific QA, just to be sure that the machine really delivers the dose that we want to deliver. We positioned the patient perfect, but we need to be sure that everything's okay.
I want to show you our approach to this. We do it in our department prior to each new SRS or SBRT case, even to each SRT case, doesn't matter. It's done in dependence of the level of complexity. It's normally done in three steps. What we do always is we do an independent monitoring unit calculation. We do a recalculation of the plan using Eclipse. As we have ARIA as our record and verify system, we have to import the plans to make them visible on the machine. If they are already there, we can easily use Eclipse to do a recalculation to compare that. Then we do portal dosimetry on the machine using the imager as a very quick and easy-to-use tool. Depending on the complexity or whenever we want to introduce maybe something newer, we do measurements of the mapped patient plan using ArcCheck SRS 1000, or even solid water phantom, EasyCube in our department, you'll see an image of that in a couple of slides. Where you can use films, TLD, iron chambers, whatever you want, everything is possible to use.
But what do we do regularly? So down here is just a little image from Eclipse. This is the calculation of the first prototype. There's some differences at that point, maybe. We do a re-calculation using Eclipse, then we print all the deviations, compare the DHS to see if everything's okay, the organs at risk are spelt correctly, and the dose given to the lesions is perfect and is okay. And then we also do a monitor unit calculation. This is done by an in-house written software. We use this already for roughly 20 years or so, and it's getting better and better. Right now the software is also able to do monitor unit calculation for VMAT treatment plans and so on. We use this for everything. It does gamma analysis. It also looks at absolute dose and everything like that, and also does a little comparison of are these monitor units in a typical range? Something like you did before. We also have a little database behind that to get an idea if something's totally screwed or not. Then we do portal dosimetry.
As I already said, we use ARIA as a record and verify system. We do a re-calculation in Eclipse, and use that re-calculation then to create a portal dose plan, and this then will be evaluated on machine. So here's just a typical image of that. You can see everything's green, everything is nice. So predicted image, actual image, and overlay of both, here's a camera analysis and a histogram, and you see a comparison of that. And then if it gets complex, or if we want to get something new and implement something new, you could also use a couple of different phantoms. Use just an image of the ArcCheck, but also the SRS 1000. The SRS array are provided by PDW, can be used easily. This is our in-house phantom, the EasyCube cubic solid water phantom with slot where if you want 16 films or TRDs or whatever. As we are already doing this for roughly 20 years. In the beginning, when we start doing SRS, there was no commercially available solutions, so we had to do everything on our own like we did our monitor unit check software on our own.
We used MATLAB to create little software routines to be able to evaluate the films and so on. Nowadays, so when we started in the first prototype, there was no Brainlab solution at that point. Nowadays, you also have an element or have a part of the software within the elements that is created for patient-specific plan QA, here's a little screenshot of that. You can import every phantom you think. In this case, it's the ArcCheck as we have in our department here. And then you can map the actual treatment plan on the patient. You can fool around with the table angles. If you deal with the ArcCheck, you want to be sure that all arcs are going on a table angle of zero, to be sure that you do not hit the electronics or something like that.
You can do that. You can override densities here. You can even fool around with the monitor units. Multiply them or so if the result is not large enough on your diodes, in this case. There's integrated solution in in the brain and in the software as well, that can be used to do this. Let's move over to the case report. We can do it. We can do the plan, we can be sure that everything's positioned, we really deployed the dose as we planned, let's see how this is working out in the patient. This is the case report of our first patient. He's 44 years old nowadays. He's a melanoma patient. The first diagnosis of his melanoma was in 2009, and then he had some surgery in 2014, end of 2014. They found lung metastases, and they resected three of them. Roughly one year later, he came to our department having brain metastases.
He was then on chemo for four cycles, but everything progressed, even the brain metastases and also metastases in the liver. And at that point we decided, okay, let's go for treating this seven lesions. In this case, you can see the image of the lesions here in the MR scans. We decided at that point to give 23Gy to all 7 lesions. As rest single shot, five table angles were used in this case. And from that day on he's on a strict regime of follow-ups. So every three months, he's going to follow-up, get an MR scan in our department or even outside. But we are still in contact with him. The last scans were done in October this year. And there were only two visible, smaller lesions. He's still on drugs since the radiotherapy, but everything is pretty nicely controlled and he's still doing good. Whenever he comes to the scans, sometimes we can't see the two lesions. Sometimes they are not even visible because they are so small. So he's doing quite good so far.
Just an idea about the overall conformity index in this case. So we got 1.61 overall of all lesions, is the volume of the brain. The V12Gy was 6.1 cc in this situation for this case, and the V3Gy, so the low dose area you can see it here was 111 cubic centimeters, cc in this situation. The next case report was a lady we treated in April this year, 48 years old female patient with breast cancer. First diagnosed in 2012, in April. In March, 2016, she came to our department because they did an MR scan, and so brain metastases in this case. So at that point we offered her whole-brain radiotherapy. We still were doing whole-brain radiotherapy at that point, and as there were two large lesions, we also did a simultaneous integrated boost to these two large legions.
Dose was 3 up to 30Gy, and the 39Gy for the simultaneous integrated boost. Exactly one year later, she came back to our department with the follow-up MR images, and we found 40 new brain metastases in this case. And as she already had whole-brain radiotherapy, and this didn't help a lot, we decided, okay, at this point, we will just use brain metastases, multiple brain metastases for this lady. We did SRS for all 40 lesions, and 13 of them got 18Gy. 18Gy, we decided to get down a little bit with the dose because of the whole-brain radiotherapy that she had upfront already. One lesion you can see down here was directly within the brainstem. So we decided, okay, we do not want to give 18Gy to this lesion, and we decided to give 12.5Gy in this lesion. And this is also nice solution or a nice possibility in the software that you can prescribe different doses to different lesions. So if you want, you can prescribe 14 different dose levels to 14 lesions. Might not make sense, but at least you can do it. In this case, we tried at least two levels. So 18Gy and 12.5Gy. To the brainstem, 5 table angles in this situation, you can see it here. We ended up with 10 arcs, so every table angle was used clockwise counterclockwise. Those distribution can be seen here. The blue area is the 3Gy isodose. So we did the cut-off here for observation at 3Gy. You see, we almost gave 3Gy to almost a whole brain, but if you look at this, you see that all the lesions are spread throughout the whole brain. This is not a big surprise that something like that happen. Here are the DVHs. You can see nicely all the DVHs for the 30 lesions that got the 18Gy prescribed. This is the one with the 12.5Gy in this case, all the organs at risk are down here. The overall conformity index, in this case, was 1.52.
So 14 lesions, all three with one single isocenter, the conformity index of this magnitude is really impressive, I think. Here's the brain, you can see that the V12Gy is 28 CC. We will just keep that in mind when we move to the following slides. So 28 CC for the V12. V3gy, as I said already, we are almost giving 3Gy to almost all the brain in this situation. The bad thing about this is that the lady came back in August with a follow-up MRI, and she had three new brain metastases in this case. So this might be small lesions that we were not able to see within the first MRI scan three months ago.
At that point, she came with three new mets. You can see them down here. We decided to treat them again because they were in the area where we didn't do too much in the first session, so we decided to treat them again. Again, 18Gy, we used 4 table angles in this situation. You can see the arrangement here. It's a little bit different to the one we had before with a little bit different angles at that point. In November, just one month ago, she had the follow-up and there were no lesions. So disease is controlled right now. At that point we are in a good stable situation. This is something that we see a lot right now, because now we are offering localized therapy to some of the lesions, and we have to do really nice follow-up every three months with imaging, just to be sure that we really can cover everything.
It happens more and more that we see patients that come back with new lesions. But the good point is, like in this case, we can just treat them again, and look what happens then, have the follow-ups and be sure that everything's covered nicely. You can see all the old lesions as well here. They are all within here because we use the same software, and we can use the fusion and map everything together and see where the old lesions are positioned to be sure when we look at the DVH so that we do not give too much dose to the old lesion. So they are all here. Within this situation they get less than 0.5Gy from this new treatment. This is just scattered then at that point.
So in this case, we got an overall conformity index of 1.45. The V12Gy is only 2.4 CC, V3Gy 64 CC, so almost nothing. And as I said, already now the disease is controlled and next follow-up will be in February also, and then we will see what happens to that. The third case report is something quite nice and something like a little bit cold in our department hyper plan or so. This is a patient 50 years old. Male patient, squamous cell lung cancer. He came into our department in July with the diagnose of six brain metastases. You can see them all here. One of these was a very, very large lesion, 3 by 2.5 by 3 centimeters roughly. We have an interdisciplinary tumor board together with the neurosurgeons and neuroradiologists and so on, and the neurosurgeons decided at that point we should go and take that out.
They did resection on the large lesion and ended up with a situation where we still have five mets within the patient and one cavity. And we then ended up with the idea of treating all these. So treating the cavity with the typical 7 times 5Gy regime hypofractionated and giving SRS to the other lesions in this situation. The nice point again, is in this situation that if you want to, as I said before, you can prescribe different doses to different lesions. In this point we ended up with three different levels of prescription dose. We prescribed 23Gy to 4 lesions, we prescribed 20Gy to 1 lesion. This was the lesion that was pretty close to the brainstem, just to be sure not to give too much dose to the brainstem in this case, so we had to use the dose.
Already 5Gy to the resection cavity. We included already the first fraction of the hypofractionated therapy within our SRS plan. This is how we deal normally with this, because this helps us already to give a well-defined dose to the cavity. If we would not include that within this treatment planning, we would end up with some completely out-of-control dose given to the resection cavity, and we need to get that back when we do the following hypofractionated regime, and need to take that into account. So we do it this way. You see 23Gy, 5Gy, so a big difference in this case. But when you look, you can see that we nicely can do this because we give the 23Gy really to this lesion in a nicely way, good conformity index in this situation.
And now this is the 5Gy isodose, the purple one. You can see, this is the cavity that we really nicely can shape the dose around the cavities. We can really give the 5Gy isotopes or give the 5Gy dose to the cavity, and give the 23Gy or 20Gy to everything else. These are the DVHs. You can see here, the 5Gy prescribed isodose or the cavity that gets 5Gy covering, the whole cavity. You know, the other lesions, the 20Gy, the 23Gy, and the overall conformity index, in this case, was 1.4. But to be honest, I excluded the cavity out of this calculation because first of all, it's a really large area, and this is not a typical situation. Conformity in this calculation doesn't make too much sense in this case, because this is nothing you will look at.
The V12Gy is large just because of this large volume of the cavity. And even the V3Gy situation is large. As I said, it's some kind of hybrid plan. The hypofractionated parts or the last six fractions for the cavity then in our department is done using the cranial SRS. The good point is we have already all contours within the elements. We can directly start and use all these contours within the same software. We have the dose there and everything, and we can take everything into account. You can see we prescribe them in this case, 30Gy because 5Gy were already given by the SRS session in this case. We did a plan, quite simple plan with only one table angle, but roughly complete arc, just not to influence too much what we already did for the SRS with the arc arrangements.
We try to be as simple as there, and not to have too much influence at that point. You can see here the DVH for the lesion at that point. And you can see down here in the cockpit, you can see that we used the old PTVs or the PTVs treated within the SRS session as organs at risk. In this case, we prescribed a strict maximum of 1Gy per lesion just to be sure to keep out all the dose out of the lesions, and not to have too much dose edit just by the hypofractionation, because we do not want to do this. If you look at the images down here, you can see the dose distributions here. You can see one of the lesions and there's one of the other lesions, and you can see that we even nicely shaped the 1Gy, which is the cutoff right now here, 1Gy around the old lesions, and we were able to spare them. So first, we can use the cavity already in the SRS session, and then we can use all the other lesions within the hypofractionated session as an alternative risk. This is how we do it in our department, and this is also something that is quite common right now.
Follow-up was, again, in November, and disease is controlled at that point. Patient is doing good and the disease is controlled. As we already did something around 100 patients, we try to think about what can we do with all this data. So we some statistics, some evaluation. At least we need to talk about risk of radionecrosis in this case, because everyone is aware of radionecrosis. We as well. So we look at all these things. But what I need to point out upfront is we do not see any increase in radionecrosis right now with our 100 patients, more than 100 patients so far. There's a 5% risk for all of these patients treated with SRS more or less. And we do see this, so their radionecrosis up to the progress, but we do not see an increase or so. That's just upfront, first of all. But when we want to talk about that, there's a nice paper out there already 2010, there are a lot of papers, but I really like this one because they did a nice study, and they also were looking at brain metastases in this case, and not on AVMs or so, so really on metastases in this case.
They were looking at radionecrosis and thought about irradiated volume as a predictor. As I said before V12Gy or V10Gy might be good predictor at that point. They really figured out that these might be really the predictors we should look at. They found out that V10Gy should be below 10.5 CC, or V12Gy if you want to look at that should be below 8 CC. If you get above that, you should think about hypofractionated treatment to lessen the risk of symptomatic radionecrosis. But as you've seen already, on my case reports, we do have patients with a V12Gy or a V10Gy that is higher than this strict number that they gave out in their paper. But we do not see an increase in radionecrosis.
That's why we thought about, radionecrosis is more or less a localized effect. It's happening surrounding the lesion you are treating. In the past, they were looking at patients that have one or two lesions, maybe three in maximum, but now we are talking about something else. So maybe we should not just look at these numbers, we should more think about having other numbers, maybe volume target, volume or something like that, or we should look at a localized V12 or V10Gy surrounding each lesion. Maybe we should end up with 14 V12Gy volumes for the lesions we treated or so. In our department, we just started plotting the V10Gy volume here for the whole-brain versus the total tumor volume to get an idea if there's some kind of correspondence or so.
Right now we found that there's a nice linear thing that you can bring in there. Right now we are thinking about, maybe we should more look at the total tumor volume and look at this, and if we are here on this and not up above this or below this, so if we are below this or on this curve, everything's fine. If we are above that, maybe we should reconsider something like that. Right now, we are just writing down V10Gy or V12Gy. We are not taking this as a strict criteria to do treatment this way or not, because we do not see any increase in that. So we're still in the process of evaluating everything, but this is already a nice start, I think.
Another thing that we are interested in is to have an idea of what about the gradient index? How good can we be, or how good should we be, at least. We also plotted the gradient index in comparison to the individual tumor volume in this case, and we figured out that most of our gradient index are between 5 and 3 for small lesions. You see if the lesion increases, the gradient index gets better and better and better. But this is just for us to have an idea, how good can we be, and if we got the plan, and it came out that we got a small lesion and or we've got a lesion of 3 CC, and we ended up with a gradient of index of 6, that we would take this plan and try to reoptimize it. Use another template, think about something else to be sure because we see we can do better at all.
Last but not least, we compared our gradient index so far and the conformity index so far that we can achieve to literature data. There's a nice paper published this year of this group here. They were also using the brain metastases element, and they were also doing VMAT plans, and did a comparison of that. I think they did the comparison of 8 plants or 10 templates, I'm not sure right now. We were in this magnitude, so we were able to achieve gradient index in the mean of 4.35, and a conformity index of 1.5. In comparison to them, we're on the same magnitude, we are even better, but I mean, our numbers are a lot larger. So if they would do it for more numbers, I think their statistics would be very, and maybe they would be in the same magnitude.
We try to be as good as this at least. In the end, to come up to a conclusion, what I wanted to point out here as our experiences that treatment of multiple brain metastases with only one isocenter is a challenging approach even for planning and delivery, as we have seen. You have to deal with this. But the good point is the Brainlab software can handle the planning process, and offering a very fast and reliable solution. The next point is due to the automatic process, and the limiting or planning variabilities in this situation. It's nice that you can really reduce the planning time and you would create very consistent, but still individualized plans for every patient. The ExacTrac system is able to compensate all possible misalignments. That's what I hopefully have shown you. The take-home message is with both systems together, patients with multiple brain metastases can easily be treated with SRS and one isocenter in Linac-based approach. So thanks for your attention. As you already have asked quite some questions within the chat, I will just briefly go through that, and try if I can answer all of these.
So the first question was, how long did it take you to commission the multiple brain metastases SRS element, and what steps did you perform? The first nice point about this is if you want to get the system running, and you're already iPlan customer, the good point is that all the data you already took and got will be the same. You will end up with the same machine profile that you already created for the iPlan. If you did all this, you should take into account for sure that you have to deal with small field sizes. You have to be sure that you have the right detector, maybe a diode, unshielded diode for the small lesions or diamond detector and so on. If you did all this, then the next step would be within the commissioning process to to see if your software behaves exactly the way that you would like to. The first point is there's an element where you can easily do the commissioning. So you can check your system by using the element part of that, by creating beams or even arcs, map them on the phantom, and then easily export them and measure them if you want to.
In our department as this element was still in development when we started with this, we created a couple of library plans that we then with single lesions, with multiple lesions, and we put them on...we mapped them on a phantom, and then measured them and recalculated them as well in Eclipse, and did the commissioning process more or less in this way. Another point is that part of the commissioning, at least if you're handling the Monte Carlo, is already done by Brainlab. This is also quite nice because you have to send all your data to them, and they will prepare the files for you for the Monte Carlo approach, and they will already end up with nice PDF document that includes a lot of statistics profiles and so on. Most of them is done by using library plans, and doing different treatment plans.
The next question is what kind of optimization parameters or structures have you used for the Eclipse plans in your plan comparison? This is a good point. I'm pretty sure that there are a lot of people out there that will be better maybe doing the treatment first, because this was already done in the past. We did it in 2014 or so, and maybe we were not that good in this case. We used ring structures surrounding the volumes to be sure to lower the dose at that point, and to use this during the planning optimization process to keep the dose conformity nicely. That's why we achieved a good conformity index, but the normal tissue sparing at that point was not that good. So we used mostly the ring structures, and then introduced some more helping structures to, which we brought in between two lesions, so they're two lesions within the same plane. I introduced another helping structure between these two lesions to try to lower the dose, the low-dose shower at that point.
It was a lot of work, but it didn't quite nicely worked out in the end, but, yeah. This is more or less what we did. The good point is, in the Brainlab solution, you do not have to think about this and deal with this. This will all be done automatically in the background. The next question would be, how long did it usually take you to generate an iPlan plan with individual isocenters? I think for this case I presented to you, for the case with the six lesions, it took me roughly two and a half to three hours. For 10 lesions, it easily increases to a full day because of all the problems that adding another isocenter influences the dose to every other isocenter. I wanted to be as strict and as accurate as possible to really only give the 20Gy I prescribed and not 22, or just because of different influences. As I had to deal with all table angles on my own, it roughly took me two and a half hours. But for the situations with 10 lesions, I think I ended up with seven hours or so. Not sure, but it took really quite long time.
When you think about the idea that you will end up with the same time on the machine, because you have to treat every single isocenter on its own, that might not be the best approach. How often do you image the ExacTrac for mets cases? What is your IGT workflow for understanding intra-fraction motions? Do you use Cone Beam CT for setting up your cranial patients too? No, we do not do Cone Beam CT. We only use ExacTrac, and ExacTrac is used in our department for all head cases. So not only the brain metastases cases, also, doesn't matter if it's a glioblastoma or whatever, we use ExacTrac for all of these patients. We do upfront imaging, positioning correction. After moving the table, we do another imaging to be sure that we applied the shifts correctly, and that the machine applied the shift correctly, and that everything is okay. At that point, we do another check, and see we are within the tolerances 0.6 millimeters and 0.6 degrees of rotation in our situation for SRS cases. The workflow is at that point, we do... Whenever we change the table angle, whenever we add a non-coplanar, move to a non-coplanar arc, we do imaging again when we reach the table position.
We do imaging, we do ExacTrac correction, we apply the shifts, we do imaging again, to be sure that shifts are applied correctly, and to be sure that we are in position again, and then we irradiate that. This is done for every single arc and every single table position. We do it for every table position. Well, image-guided radiation workflow for understanding infra-fraction motion. It's more or less like this. Whenever there's something, we would see that, but we do not see any real. Within the brain or within the skull, we do not see any motion. There shouldn't be any motion at all for brain metastases. So we do not see anything like that. But as I said already, we do imaging upfront before we treat every single table angle or every single arc to be sure that still everything is in place because we are using a mask system, and there might be room for the patient to move within the mask.
What clinical margins do you place for PTV definition for multiple brain metastases cases? And, is this a uniform CTV expansion? What margins have you used for iPlan? In the past we used margins for iPlan of 1 or 2 millimeters depending on how good the lesion was visual. If it's a cystic lesion, and the borders of the lesion cannot be clearly defined or so, we end up with a 2-millimeter margin, same for elements as well. But normally we only apply 1-millimeter uniform margin most of the time. If we are pretty sure, imaging is really nice and new, and MRI scans were done just a day before treatment starts or whatever. We also end up in some situations with a margin of zero depending on how good we can really visualize and define the lesion. Margins are only there for compensating for problems during delineation of the lesion, not for the table or so on, because we are within the 0.6 millimeters for all of that.
There's another question. How do you verify the rotational accuracy? And then in particular the rotation of calibration of ExacTrac. I'm not quite sure what you mean by rotational calibration of the ExacTrac. But we do verify the ExacTrac by using the software or the in-house or the tools that were offered already by Brainlab. There are a couple of phantoms and a couple of tools that has to be used. You have to check the radiation isocenter or the calibration of the radiation isocenter on a regular basis for the software. We do Winston-Lutz test upfront. This has done or for all cases upfront. And so we do this full-day end-to-end test on a regular basis that really includes also positioning of the patient using the ExacTraC in the full workflow.
So we have the BB, and then we have the phantom there with an iron chamber and also with films. And then we irradiate that, and see if everything's in place. We do not do that once per week or so, but we do it on a regular basis at least every three month to check that everything is okay at that point. Hopefully, I answered your question at least a little bit.The next question would be for patient-specific quality assurance, what is your passing rate when comparing the monitor units with either Eclipse or your in-house software? The passing rate for portal dose is two and two, in this case, we use two and two. The passing rate for comparing the monitor units with our in-house software is also 2% at that point.
For Eclipse, it's a little bit more complicated because the situation in our institution is that our Eclipse is not commissioned for a pencil beam. We did AAA and Acuros as there is another question down below here in the questions. There might be already a difference just because of a different dose algorithm. At that point we say, okay, for lesions that are more or less central located within the brain, we are within 2%. But for lesions that are more next to the skull or so that are more next to the inhomogeneities, we can also end up with 3% or 5%. The recalculation is just to see if there's really something very, very strange. But most of the time we end up with roughly in the magnitude of 3%, even if we use a AAA or not pencil beam at that point.
Another question was, for your multiple mets plans and elements, how many of the plans required repeating gantry arcs to cover all lesions? Also if gantry arcs have to be repeated, how could the delivery time be comparable to VMAT RapidArc? The delivery time... It depends a bit. So I've no number really for how many of the plans require repeating gantry angles. Repeating gantry angles is most of the time more effect of giving enough dose to some of the lesions, as you said, in your question, or if you have a lot of lesions that are all within the same plane, you were doing a clockwise rotation, and you already doing three of the lesions, and doing a counterclockwise rotation, and irradiating another three and so or two lesions and not the same, maybe. One of the processes within the optimization process is to cover as much lesions as possible with as many RX as possible to streamline the process.
The amount or the treatment time can still be comparable to VMAT due to the fact for the RapidArc solution, you do a volumated...so an IMT approach. A lot of your monitor units were just going within the MLC or within your drawers without even adding any dose to the patient. When you do VMAT, your dose will be... At some point, the dose rate will go down from maybe if you have 600 monitor units as default, will go to 300 or something like that, while doing dynamic conformal RX, your dose rate will be stable at 600 monitor units. You will still always give 600 monitor units. In the end, dose and the patient is directly dependent to monitor units. You end up with the same amount of monitoring units or roughly the same, so you end up with the same at least in the same magnitude of delivery time.
The next question was something that I already answered a little bit maybe, but I would just go to that. Both AAA and Acruos were shown to have issues for calculation when targets are small, roughly 1 cm. There's the paper for [inaudible 01:16:15] medical physics, 2011. Do you see systematic differences from elements multiple brain metastases when you do your check in Eclipse? What are your smallest fields measured for Eclipse? As I said already there are differences. This might be also depending on that effect, but our...the good point is at the same time when we did the commissioning for the iPlan and the elements or iPlan, to be honest, because we started with iPlan, we also did the commissioning for the Eclipse system, because it also came new into our department at that point, and we used the same more or less. We included measurements in the Eclipse down to 1 centimeter by square field size for our protectors and so on.
We measured that to quite small field sizes, not just down to 3 by 3 or so, we went down to 1 by 1 as we did the measurements already for the iPlan, and they were there. But in the iPlan, we go down to 0.5 by 0.5. These are the smallest fields that we had in the Eclipse. That's why our results might be quite nicely, maybe. Why do you monitor 3Gy for your patients in addition to 12Gy and 10Gy? We do this just to have an idea what the low dose is given to the patient, just to have an idea how much low dose we were giving in terms of... This is a little bit like a quality parameter for all our treatment planning.
We do not want to give too much low dose to everything else. That's why we monitor. There's no real clinical reason right now, or something like that. This is just as a parameter to monitor, to have an idea, how much low dose, what about the low dose spillage that we were giving to the patient? In addition to the other ones that we still think about radionecrosis also, this has no directly link to clinical data so far, but it's a parameter for us just to have an idea of plan quality. Because we want to keep this as low as possible because there do not need to be so much dose in the normal brain.
Next question would be, how many templates do you use clinical, and how do they differ? That's a really good question, because I think when we started with the prototype, we started with roughly 30 templates or something like that, because upfront, we thought about all situations, and we created templates for almost every situation, and about everything that we would think would come into our department. Right now I think, we ended up with eight or nine. We got templates for situations where we would only use one template. 5RX equally split more or less around the skull. And then we have templates for situations where we only would like to treat more on the left or more on the right side with RX more on that side. And then we have templates that have different numbers of minimal used arcs. One point within the templates, or one point within the optimization process is that the software looks at the numbers of arcs used, or that needs to be used.
We found out that we had situations where the system came up with a solution that three arcs would be sufficient to give the dose to the patient. You end up with a nice conformer plan, but we thought about maybe it would be better to end up with 4RX or even 5RX, and then we have different situations where we forced the system to use at least 4RX or 5RX. This makes another... So you have to have every other templates multiplied by this, so we end up with nine or so. But we mostly use the 5-arc template, and give the software as much freedom as possible to come up with 3RX, if necessary, and if that's enough. It's pretty easy. You do not need to have
Manuel: Thank you, Bocton, for this nice introduction. I'm happy to share my, or our experience, the European experience or more the Hamburg experience with the elements multiple brain mets SRS.
Bocton: Thank you.
Manuel: What are we going to talk about within the next couple of minutes, next hour, roughly. First of all, I would like to give you a short introduction about our department a little bit, so that you have an idea of what our department looks like, what kind of patients we do have and so on. And then a little outlook on the history of the multiple brain metastases element in our department. At which point did we get involved into development maybe and so on, and just to give you an idea why I want to talk about that. And then my talk is more or less split into three different parts. So the first one would be that I wanted to talk about what is needed to get theoretically a good dose distribution.
How can treatment planning be done more or less? In that terms, I would like to talk about what kind of patients are suitable, what kind of different options are there for these patients. We will have a quick, roughly look at two papers that are out there that might help us in decision-making. And then I would like to compare some kind of different treatment options that are out there for patients with multiple brain metastases. I just wanted to get you through the whole process that we did in our department when we got in touch with the software, and we thought about what is the purpose of the software, is it good, how can we deal with this and so on. If we can get theoretically a good dose distribution, then I would like to talk about how can we realize this dose distribution for the patient.
And then we have to talk about uncertainties little bit about image guidance. This is a very, very important part because as you know, we deal with one virtual ISO center, and at that point, we need to be sure that image guidance is working quite nicely and that we really hit what we want to hit. If we have done all the image guidance, then it's up to patient-specific QA. So how to get the dose really to the patient, and does the machine work properly and everything like that. I will tell a bit about the way we were doing that and how we can do it with the element solution. In the last part, I would like to give you some case reports of patients that we treated in our department here, and to give you an idea how everything would look like. Three or four patients I would like to present, and then a little bit some clinical results and a short outlook. That should be everything I want to present to you.
First of all, some key facts about our department. Down here is our Novalis STx system that is used for treating all these patients with brain tumors and brain metastases. We do have approximately 2000 new patients per year, and roughly 200 of them are really brain metastases patients. And we have another 100 brain tumor patients. These two indications are within the top five, that are treated in our department. So that's why we decided to use brain lab software, because we really, really believe in the technology. We are a Brainlab customer since 2000. As I said before, we have a Novalis STx system with the ExacTrac, the 60 robotics couch that is used to treat these patients. What about Multi Brain Metastases SRS at the UKE at our university medical center.
So, we first came in touch with Brainlab in the end of 2013. So roughly four years ago. They were here, and we had a usability evaluation of the really first prototype of the software. So they came up with the idea of treating multiple brain metastases with a virtual ISO center, and they asked us, what do you think about that? We tried it, we had some nice ideas, nice thoughts, and then we figured out how to get a research and development program together with Brainlab. In April, 2014, and then the first prototype was installed in our department. We were really doing a lot of development and research with that. And then another prototype was installed in May, 2015...on April, 2015. We treated our first patient using this prototype. The beginning 2017, we were the first European installation of the released clinical system. Right now, we are using the Cranial SRS, the Brainlab Multi Brain Metastases SRS, and also the Spine SRS software were installed in our department, and are clinically used.
We have more than 100 patients treated using the elements and the Brain Metastases SRS and over 20 patients treated using Cranial SRS, and numbers are increasing constantly. This is all about our department, so who should be treated. I think all of you know out there that in the past whole-brain radiotherapy was more or less the option that Linac-based clinics would offer patients with multiple brain metastases. Multiple brain metastases in the past meant most of the time, more than four or more than five lesions. So it was more like you got a diagnostic department, and they were counting the lesions and it was more like one, two, three, four, five, multiple. And whenever you got the diagnosis of multiple brain metastases, you used whole-brain radiation, that was in the past days.
But today I think, we all want to do it in a more local way by using SRS just because of preserving neurocognitive function and so on. Unfortunately in the past, the possibilities for this treatment were very limited. I mean, there was Gamma Knife out there for decades that were doing this as well, but there was no nice Linac-based approach for this. Just to give you an idea that brain mets is still still a hot topic. When I prepared this talk, I just did a little research on Google Scholar, and I just entered multiple brain metastases one isocenter, you can see it here. I just wanted to know how much papers roughly were published within 2016 and 2017 now. It came up with roughly 820 papers.
Brain metastases using one single-isocenter is really a hot topic, and it's really of interest out there, just to give you an idea that we are not talking about something that might be already not on vocal or anything like that. So what were the three major treatment options? I mean, among others, there are systemic treatments, chemo, and so on. But what are the major three major options? In the past, it was whole-brain radiotherapy, maybe with or without hippocampal sparing. We've got neurosurgery of single individual mets, then maybe combined with whole-brain radiotherapy or nowadays combined with SRS. I will give you a case within the case reports of that. SRS or STx, so hypofractionated therapy or single-fraction for multiple brain metastases alone. This could be combined with whole-brain radiotherapy if you want or not, but just to give something upfront, whole-brain radiotherapy will not help you at all. So there's no reason to implement that at all.
So what we thought about in our department is that everyone knows that whole-brain radiotherapy can be a toxic treatment. And the question is, do the benefits justify the risk in this situation? Therefore it was quite nice that Brainlab came to us with this nice idea of software. We, from day one on, were really interested and see how it works and how it could be used. What we did then, but in the past, we did a little study on papers out there just to have an idea of what we should do and how we could deal with this, and how to decide, which would be the best solution to deal with this disease for this patient to help a patient the best.
Whenever you talk about that, about multiple brain metastases, I think the first study that will come up will be the Yamamoto study. I will talk about that in the next minutes. Published already in 2014, there was an update just some months ago, but still the results are the same. Then there's a Kocher study in Germany already 2010, and there are a couple of other studies, but just focus first on the Yamamoto paper. This is a really nice paper because it's a multi-institutional study, and it really has roughly 1,200 patients in there. It's a really, really large study. All of these patients were treated with Gamma Knife. So this is a Gamma Knife study. They divided their patients into three groups. So one group was patients with only one single met, and the second group was patients with 2 to 4 mets, and the third group was patients with 5 to 10 mets, so something, what we would call multiple brain metastases in the past.
The results were that with 5 to 10 brain metastases, the treatment is non-inferior or the outcome for the patient is non-inferior to that with patients with two to four brain metastases. Two to four or two to three would be something that we already treated with a Linac-based approach in the past. This is something that we already knew that we can do good results on that. So very promising. It doesn't make sense to split groups there. So we can treat patients with up to 10 brain metastases, and our Gamma Knife community is still, already does that for 25 or so lesions in a row if they want to.
Just some key facts out of the study. The first one is that there were no grade 5, and almost no grade 4 adverse events during this study. Doesn't matter if it was 1 tumor, 2 to 4 or 5 to 10 tumors, all goes with the same. The next very important point is that the maintain...Kocher function was also the same for all of these groups. It was really, really nice. And you see, we got here numbers of 94% and 95% already for four months after SRS. The neurocognitive function was very nicely preserved for all of these patients. The last, maybe most important information is that if you look at the overall survival, you will see here, the blue and the red curve are for the 2 to 4, and 5 to 10 mets groups, and you see they are the same. So there's no change in overall survival for the patients. Doesn't matter if they got 2 lesions or got 10 lesions, it doesn't matter. They are all behaving the same. They have the overall survival curve, and there is no difference.
The one tumor group is a complete other story. So they are doing better, but this is more or less complete different disease here, because they are doing better in all other points of perspectives as well. So then just a quick look at another study, because I think this is a really nice paper because the idea behind this paper is does the number of tumors matter? It's a little bit like, do we really need to talk about number of lesions or do we have to talk about something else? They found out that the number is a really nice surrogate for the disease, but it's not an independent predictive factor. It's more like a surrogate for a lot of other clinical factors. So people with more lesions are doing worse just because of other diseases and other problems that they have as well. And then they came up with the idea that for patients with more than four brain metastases, SRS could be a good approach. It should be the future approach at all.
All these things convinced us already that we should go this way and follow the Brainlab road, and get more into detail with this. What we then did at that point is we used the prototype of the software and compared it to three other possibilities to do treatment planning and treatment for these patients, just because we wanted to know how good the software really is in comparison to the other things. We compared it to iPlan. So this would be an approach, multiple isocenter dynamic conformal arc, like we did in the past. For one, two, three lesions, we use single isocenter for every lesion, then use a couple of RX, or a couple of beams, and try to figure out, and try to arrange a good dose distribution at that point. We just try to do it for patients with up to 10 mets. So we use this on 16 different datasets.
We use Eclipse volumated arc therapy at that point, so RapidArc approach. We have another institution here in Hamburg that had had a Gammaknife, and that were offering us the possibility to do treatment planning for all these patients using the Gammaknife, because this was more or less the gold standard in the past, because they were doing this already for 20 years or so. So how did we do it? We did pre-contour all the datasets in Eclipse using...in elements using the automatic fusion and using all the nice algorithms and tools for outlining and delineation of the targets. And then we entered this pre-contour datasets into the three other treatment planning modalities, and in elements, we used one isocenter. So the virtual isocenter approach that will be positioned in the center of mass of all lesions. We used a template with up to 5 table angles, and that came up to 10 arcs in total, and this wasn't completely automatic approach. So just hit the calculate button and the software came up with a solution within three to five minutes.
For iPlan, as I said, the multi-isocenter dynamic conformal arc approach, we used one isocenter per lesion, three table angles with up to five arcs in total. This is a complete manual approach. So you have to do it all-forward planning. So start with the first lesion, arrange everything, at the next lesion, arrange everything, and just have to fool around a lot with table angles and so on to be sure that you do not influence each other too much, because that has to be taken into account as well. For the Eclipse approach, we used the same isocenter that the element came up with, we used 5 table angles as well, and also 10 arcs as well in total. So we used exactly the same setup, like brain metastases already came up, like the element already came up with. The only thing that we changed was we changed the column data field opening to allow the Eclipse software to cover all lesions within one arc if it thought it would be needful. So just to get more degrees of freedom for the software to be more convenient there.
For the Gammaknife, they used up to 10 shots. Was done with Leksell GammaPlan, up to 10 shots per lesion with maximum of 25 minutes beam on time per lesion. Let's have a look at the results. The next slides will all look the same more or less. So we will have the brain metastases on the left side, the dose distribution for that. The other solution on the other side that we would like to compare with, we got a little tabular down here to have an idea about the V12Gy. So something that might be V12 or V10Gy, that might be a good predictor for radionecrosis or so. So the volume of the brain receiving 12Gy, and we got the volume of the brain receiving 3Gy to get an idea about low-dose shower, or low-dose spillage at all.
This is just for one of the cases, to give an example. So it was six metastases cases, 1.5 CC tumor volume in total, and we prescribed 20Gy to the isodose covering the lesion. What you can see on the first look is that the brain metastases case is really nicely conformal. The blue area, the dose cut-off is at 3Gy in all cases. You see that the low dose shower is a lot lower than for the VMAT approach in this case. Even the V12Gy is half the amount of that, what came up to the VMAT approach. What you can see, it's nicely conformal. You can see already on the VMAT approach, that there will be another lesion in both situations here.
Another lesion, just a little bit ahead...more cranial to the cranial direction that the brain metastases solution can really spare and can really split the dose distribution to both. You already see that you get a high dose area here for the VMAT approach, so it's not the conformity at all compared to the other one. Same, if you look at the other dimension, you see the large, low dose shower here. You also see some three gray areas over here where there's not even in lesion in this case, in this situation. If we move over to the multiple isocenter dynamic conformal arc approach, you'll see that the dose distribution is also nicely conformal. You see no high dose areas here next to it, maybe a little bit here, but not here, next to it. The V12Gy is in the same magnitude and even the V3Gy.
So the low dose is only 200 CC compared to 700 something that we have for the VMAT. Even the low dose shower is a little bit lower. This also looks like a nice approach for doing treatment planning despite the fact that it's really time-consuming because you have to do it all-forward planning and have to spend a lot of time on that, and you will also need to spend a lot of time on the machine, but I will come to that in a couple of seconds here within the next slides. So good approach. Next comparison would be the Gammaknife. Sorry for having not the same visualization for the Gammaknife, but this is just due to the fact that the GammaPlan software is not able to export the dose distribution in a good way so that you can import it into any other software, Eclipse, for instance, what I used here.
In this case, I only pointed out here the 3Gy isodose line to give you an idea about the low-dose shower. We were only able to say that V12Gy and 3Gy are in the same magnitude for the GammaPlan and for the Gammaknife situation. I didn't get even numbers here. So because of the fact that the Gammaplan looks at separate dose distribution or dose cubes around each lesion and does not really offer a complete dose distribution or dose volume for the whole brain that is calculated really completely in detail. So we just end up with the same magnitude. To put that all together, I created a little comparison chart, or we created a little comparison chart in our department. We looked in this comparison at the conformity index. So this is inverse Paddick. It should be one to be perfect, same for the gradient index. We looked at the dose spillage in terms of the volume received in 3Gy.
We looked at the planning time and even at the delivery time and monitor units where applicable. Up here are the brain metastases numbers. You see, you can get a nice plan within 5 to 10 minutes, depending on if you have to arrange something a little...change an arrangement a little bit and hit the calculate button again. So 5 times 5 to 10 minutes was what the prototype took right now with a clinical system, we are more in the magnitude of 5 minutes or so for our plan. And it's even calculating a little bit faster. Delivery time is really interesting. As we only deal with one isocenter, all lesions were treated with the same isocenter.
The delivery time is only dependent on the prescription. So if you prescribe 23Gy, it takes a little bit longer than prescribing 18Gy or so. But you only have to apply that once. This is the biggest drawback for the multiple isocenter dynamic conformal arc approach. Because at that point, if you want to treat 10 lesions, you have to apply 10 times 18 or 23Gy. So you end up with this time multiplied by the number of lesions, and this will easily end up for 10 lesions with five hours or so, and this will really then kill all your schedule, and you will only be able to treat on patients. So that's one of the points and one of the reasons why in the past, no one using linear accelerator thought about treating multiple lesions within one session. There was no real solution for that.
The VMAT was...delivery time was the same, conformity index was also in the same magnitude, but the gradient index, so the dose fall off was worse, and the dose spillage as we have seen on the images was also very bad compared to the brain metastases case in my situation here, at least. Planning time was even longer because it was not an automatic process. Just calculating dose for dose grid of one millimeter just roughly takes 20 minutes or so. So just calculating is already two times the planning time for the whole-brain mestatases approach. Gammaknife, you see it's the same dose spillage might be the same, even better...I cannot really say something about that because the numbers were not in a good way prepared at that point. What I can say is the delivery time is worse because for 10 lesions, it takes you roughly two and a half up to three hours treatment time for the patients or beam on time on the table. The good point is the machine doesn't make any sound. So the patient can fall asleep during this time, but still the machine is occupied by the patient for time.
If we can do and provide good dose distributions theoretically, so we can use elements to get a nicely conformal dose plan for the patient. And we have to think about what comes if we move to the machine. We got a lot of non-coplanar RX. In the past, when we were dealing with non-coplanar RX, and we used the multiple isocenter approach, we had a lesion directly in the isocenter. Whenever we got a rotational error, and this can easily happen because the couch rotation is the weakest part in the whole chain and the whole treatment chain. When we did this in the past, that wasn't a very big deal because the lesion was in the isocenter, and if we introduce an error rotating around the isocenter, the resulting error was not that high.
Maybe we ended up adding a little margin as well. All the lesions were round-shaped. So that was not a real big deal or real big problem. But nowadays, if you use something like virtual isocenter, if you look at these, you see that there's no lesion in the isocenter at all. These are all the lesions in this case, the yellow, the orange, and the blue one, you see the leaf of field openings here were treated for this arc. And then just to get an idea of what happens, I introduced a little rotational error. I fooled around a little bit with the DICOM data and introduced a rotational error. What you can see is that the yellow, the orange and the blue one, they all move out of the field opening. You will miss these lesions. To come up with some numbers, I just did a calculation then of the plan.
I introduced translational errors here. You can see that the mean dose decreases by 2.5Gy in this case, and the minimum doses decreases almost by 7Gy if you got a 2-millimeter translation error and it already starts for one millimeter to really create problems. The same for rotational error. Rotational errors can really easily be introduced. And rotational error of one degree in pitch and roll, if you do know image correction or something like that or imaging, this will happen and you will not see it at all. You will end up with an under dosage here. Not even an under dosage for the lesion as well, you will end up with an overdosage to the normal brain because you will not hit the lesion, but the normal brain. It's getting worse if the distance of the lesion to the isocenter to the virtual isocenter increases, because it's just a geometrical problem. Just mathematics behind that that end up with this.
We need to have something to deal with this. The good point is we have nowadays frameless radiosurgery. There's a nice paper out there since 2011 by the Brussels Group by Thierry Gevaert who compared the frame-based approach, so with an head ring to the frameless approach. They ended up that frameless is even better than frame-based, just because of the fact that you introduce imaging. For the frame-based approach, you have no imaging. You rely completely on the fact that the patient is completely immobilized. If you move over to frameless, you have a mask, but you do imaging, do correction for that. At that point, you get even better than the frame-based approach. In their publication, they ended up with an error of roughly 0.7 millimeters in the 3D vector for all three orientation.
Frameless is a good approach and we can use it. How do we do it? We do it with ExacTrac. What I really recommend is whenever you change the table angle, and move over to a non-coplanar arc table angle, you should do imaging using ExacTrac. You should apply the corrections, and then you should do another set of images just to be sure that the correction was applied correctly, and then start treating that table angle again. They ended up with this 0.7 millimeters, and this is also just mathematics. This more or less depends on the imaging that you use for treatment planning. On the resolution of the CT, because it's based on the TRR images that are used in the ExacTrac. If you got a better resolution for the CT, for image planning, for treatment planning, you will end up with a bad result for the positioning because your TRRs will not be exactly in the Exac. At least you should use images of one-millimeter resolution, at least.
What can we do for quality assurance for this? We all know we have the Winston-Lutz test, which is used for the lasers mounted to find if they really are hitting the radiation isocenter defined by the linear accelerator, but now we introduce another system, the ExacTrac system, and this has its own imaging isocenter. This is created by the x-ray imaging and the infrared marker system. If you want to be exact and accurate, you need to be aware that your x-ray isocenter is exactly the same like your actual radiation isocenter of the accelerator. A good tool to check that is maybe a hidden target test or an end-to-end test. Just to give you a short idea, how we do it in our department, we use the STEEV Phantom here. We have a little BB in there, you can see it here. So from some of the CT slices, and we do the imaging, then we do the planning using the element. Like you can see here, we do the localization using the ExacTrac. We do the treatment delivery. At that point, we have an iron chamber or films inside of our Phantom. You do an analysis of that. In the end, we want to end up with one-millimeter end-to-end localization, and 2% dosimetric evaluation. This is something that we can achieve, and this is something that we do at least on a regular basis. But yeah, it can be done in this way.
Another point that I really would want to bring up is this really nice publication that I really would recommend to all of you out there. It's done by Tominaga Hirofumi, published already in 2014. They had an ExacTrac system and then Novalis STx, and they wanted to figure out if the ExacTrac system is really able for compensate for all these errors. This is something that we are also interested in. If we want to deal with this, we need a system that really is able to compensate for the rotational errors. Their approach was quite nice and quite easy. This is really beautiful. They created the little cubic Phantom with a couple of spheres in it. One was the target in the isocenter. This was the reference target. They introduced three other spheres here around this, these blue ones here. They were used for image-guided marker fusion, so that you could use the ExacTrac and also for Cone Beam CT. And then they just brought in some targets.
Four targets at 52 millimeters distance from the isocenter, these would be covered by the 2.5-millimeter leaves of the STx system, and another four targets at 87 millimeters distance, this would be covered by the 5-millimeter leaves in this case, so they could use the Phantom forces. Other things, I just want to look at the idea of if the patient positioning or the positioning accuracy can really be achieved by using this. What they did then is they positioned their Phantom correctly, did a Cone Beam CT, did ExacTrac, used the Cone Beam CT then as the gold standard, and then introduced errors. First of all, they introduced 0.5-degree error in roll and pitch. The Cone Beam CT at that point marker fusion, and ended up with a deviation of already 1 millimeter in the anterior-posterior direction. This one here.
And then they said, okay, but we can have 1 degree in roll and pitch as well. So 0.5 would be perfect, but 1 degree is really easily possible and can happen. They introduced that, did again a Cone Beam CT, and they ended already up with 2.3 millimeter in the anterior-posterior direction. If you look... Remember my DVH images that I showed you, you will see that this will already end up with an unbelievable under dosage to your lesion, and you will really miss what you wanted to hit at that point. What they then did is they did ExacTrac, did automatic fusion, did correction, rearranged the table position, did another ExacTrac, and did a Cone Beam CT, and then at that point, just a Cone Beam CT, just to see if everything is really in place, because that was the gold standard at that point, and they ended up with a maximum deviation of 0.6 millimeters.
ExacTrac was really able to compensate for all these things, put everything back to 0.6 millimeters, which is absolutely perfect within stereotactic dimension so everything should perfectly work, and is a nice tool. This is why you really should use this, and I really recommend to do imaging whenever you change the table here. So now we can theoretically do doses distribution by using elements. We have ExacTrac onboard, so we are pretty sure that we really hit what we want to hit. The last point in the whole chain would be patient-specific QA, just to be sure that the machine really delivers the dose that we want to deliver. We positioned the patient perfect, but we need to be sure that everything's okay.
I want to show you our approach to this. We do it in our department prior to each new SRS or SBRT case, even to each SRT case, doesn't matter. It's done in dependence of the level of complexity. It's normally done in three steps. What we do always is we do an independent monitoring unit calculation. We do a recalculation of the plan using Eclipse. As we have ARIA as our record and verify system, we have to import the plans to make them visible on the machine. If they are already there, we can easily use Eclipse to do a recalculation to compare that. Then we do portal dosimetry on the machine using the imager as a very quick and easy-to-use tool. Depending on the complexity or whenever we want to introduce maybe something newer, we do measurements of the mapped patient plan using ArcCheck SRS 1000, or even solid water phantom, EasyCube in our department, you'll see an image of that in a couple of slides. Where you can use films, TLD, iron chambers, whatever you want, everything is possible to use.
But what do we do regularly? So down here is just a little image from Eclipse. This is the calculation of the first prototype. There's some differences at that point, maybe. We do a re-calculation using Eclipse, then we print all the deviations, compare the DHS to see if everything's okay, the organs at risk are spelt correctly, and the dose given to the lesions is perfect and is okay. And then we also do a monitor unit calculation. This is done by an in-house written software. We use this already for roughly 20 years or so, and it's getting better and better. Right now the software is also able to do monitor unit calculation for VMAT treatment plans and so on. We use this for everything. It does gamma analysis. It also looks at absolute dose and everything like that, and also does a little comparison of are these monitor units in a typical range? Something like you did before. We also have a little database behind that to get an idea if something's totally screwed or not. Then we do portal dosimetry.
As I already said, we use ARIA as a record and verify system. We do a re-calculation in Eclipse, and use that re-calculation then to create a portal dose plan, and this then will be evaluated on machine. So here's just a typical image of that. You can see everything's green, everything is nice. So predicted image, actual image, and overlay of both, here's a camera analysis and a histogram, and you see a comparison of that. And then if it gets complex, or if we want to get something new and implement something new, you could also use a couple of different phantoms. Use just an image of the ArcCheck, but also the SRS 1000. The SRS array are provided by PDW, can be used easily. This is our in-house phantom, the EasyCube cubic solid water phantom with slot where if you want 16 films or TRDs or whatever. As we are already doing this for roughly 20 years. In the beginning, when we start doing SRS, there was no commercially available solutions, so we had to do everything on our own like we did our monitor unit check software on our own.
We used MATLAB to create little software routines to be able to evaluate the films and so on. Nowadays, so when we started in the first prototype, there was no Brainlab solution at that point. Nowadays, you also have an element or have a part of the software within the elements that is created for patient-specific plan QA, here's a little screenshot of that. You can import every phantom you think. In this case, it's the ArcCheck as we have in our department here. And then you can map the actual treatment plan on the patient. You can fool around with the table angles. If you deal with the ArcCheck, you want to be sure that all arcs are going on a table angle of zero, to be sure that you do not hit the electronics or something like that.
You can do that. You can override densities here. You can even fool around with the monitor units. Multiply them or so if the result is not large enough on your diodes, in this case. There's integrated solution in in the brain and in the software as well, that can be used to do this. Let's move over to the case report. We can do it. We can do the plan, we can be sure that everything's positioned, we really deployed the dose as we planned, let's see how this is working out in the patient. This is the case report of our first patient. He's 44 years old nowadays. He's a melanoma patient. The first diagnosis of his melanoma was in 2009, and then he had some surgery in 2014, end of 2014. They found lung metastases, and they resected three of them. Roughly one year later, he came to our department having brain metastases.
He was then on chemo for four cycles, but everything progressed, even the brain metastases and also metastases in the liver. And at that point we decided, okay, let's go for treating this seven lesions. In this case, you can see the image of the lesions here in the MR scans. We decided at that point to give 23Gy to all 7 lesions. As rest single shot, five table angles were used in this case. And from that day on he's on a strict regime of follow-ups. So every three months, he's going to follow-up, get an MR scan in our department or even outside. But we are still in contact with him. The last scans were done in October this year. And there were only two visible, smaller lesions. He's still on drugs since the radiotherapy, but everything is pretty nicely controlled and he's still doing good. Whenever he comes to the scans, sometimes we can't see the two lesions. Sometimes they are not even visible because they are so small. So he's doing quite good so far.
Just an idea about the overall conformity index in this case. So we got 1.61 overall of all lesions, is the volume of the brain. The V12Gy was 6.1 cc in this situation for this case, and the V3Gy, so the low dose area you can see it here was 111 cubic centimeters, cc in this situation. The next case report was a lady we treated in April this year, 48 years old female patient with breast cancer. First diagnosed in 2012, in April. In March, 2016, she came to our department because they did an MR scan, and so brain metastases in this case. So at that point we offered her whole-brain radiotherapy. We still were doing whole-brain radiotherapy at that point, and as there were two large lesions, we also did a simultaneous integrated boost to these two large legions.
Dose was 3 up to 30Gy, and the 39Gy for the simultaneous integrated boost. Exactly one year later, she came back to our department with the follow-up MR images, and we found 40 new brain metastases in this case. And as she already had whole-brain radiotherapy, and this didn't help a lot, we decided, okay, at this point, we will just use brain metastases, multiple brain metastases for this lady. We did SRS for all 40 lesions, and 13 of them got 18Gy. 18Gy, we decided to get down a little bit with the dose because of the whole-brain radiotherapy that she had upfront already. One lesion you can see down here was directly within the brainstem. So we decided, okay, we do not want to give 18Gy to this lesion, and we decided to give 12.5Gy in this lesion. And this is also nice solution or a nice possibility in the software that you can prescribe different doses to different lesions. So if you want, you can prescribe 14 different dose levels to 14 lesions. Might not make sense, but at least you can do it. In this case, we tried at least two levels. So 18Gy and 12.5Gy. To the brainstem, 5 table angles in this situation, you can see it here. We ended up with 10 arcs, so every table angle was used clockwise counterclockwise. Those distribution can be seen here. The blue area is the 3Gy isodose. So we did the cut-off here for observation at 3Gy. You see, we almost gave 3Gy to almost a whole brain, but if you look at this, you see that all the lesions are spread throughout the whole brain. This is not a big surprise that something like that happen. Here are the DVHs. You can see nicely all the DVHs for the 30 lesions that got the 18Gy prescribed. This is the one with the 12.5Gy in this case, all the organs at risk are down here. The overall conformity index, in this case, was 1.52.
So 14 lesions, all three with one single isocenter, the conformity index of this magnitude is really impressive, I think. Here's the brain, you can see that the V12Gy is 28 CC. We will just keep that in mind when we move to the following slides. So 28 CC for the V12. V3gy, as I said already, we are almost giving 3Gy to almost all the brain in this situation. The bad thing about this is that the lady came back in August with a follow-up MRI, and she had three new brain metastases in this case. So this might be small lesions that we were not able to see within the first MRI scan three months ago.
At that point, she came with three new mets. You can see them down here. We decided to treat them again because they were in the area where we didn't do too much in the first session, so we decided to treat them again. Again, 18Gy, we used 4 table angles in this situation. You can see the arrangement here. It's a little bit different to the one we had before with a little bit different angles at that point. In November, just one month ago, she had the follow-up and there were no lesions. So disease is controlled right now. At that point we are in a good stable situation. This is something that we see a lot right now, because now we are offering localized therapy to some of the lesions, and we have to do really nice follow-up every three months with imaging, just to be sure that we really can cover everything.
It happens more and more that we see patients that come back with new lesions. But the good point is, like in this case, we can just treat them again, and look what happens then, have the follow-ups and be sure that everything's covered nicely. You can see all the old lesions as well here. They are all within here because we use the same software, and we can use the fusion and map everything together and see where the old lesions are positioned to be sure when we look at the DVH so that we do not give too much dose to the old lesion. So they are all here. Within this situation they get less than 0.5Gy from this new treatment. This is just scattered then at that point.
So in this case, we got an overall conformity index of 1.45. The V12Gy is only 2.4 CC, V3Gy 64 CC, so almost nothing. And as I said, already now the disease is controlled and next follow-up will be in February also, and then we will see what happens to that. The third case report is something quite nice and something like a little bit cold in our department hyper plan or so. This is a patient 50 years old. Male patient, squamous cell lung cancer. He came into our department in July with the diagnose of six brain metastases. You can see them all here. One of these was a very, very large lesion, 3 by 2.5 by 3 centimeters roughly. We have an interdisciplinary tumor board together with the neurosurgeons and neuroradiologists and so on, and the neurosurgeons decided at that point we should go and take that out.
They did resection on the large lesion and ended up with a situation where we still have five mets within the patient and one cavity. And we then ended up with the idea of treating all these. So treating the cavity with the typical 7 times 5Gy regime hypofractionated and giving SRS to the other lesions in this situation. The nice point again, is in this situation that if you want to, as I said before, you can prescribe different doses to different lesions. In this point we ended up with three different levels of prescription dose. We prescribed 23Gy to 4 lesions, we prescribed 20Gy to 1 lesion. This was the lesion that was pretty close to the brainstem, just to be sure not to give too much dose to the brainstem in this case, so we had to use the dose.
Already 5Gy to the resection cavity. We included already the first fraction of the hypofractionated therapy within our SRS plan. This is how we deal normally with this, because this helps us already to give a well-defined dose to the cavity. If we would not include that within this treatment planning, we would end up with some completely out-of-control dose given to the resection cavity, and we need to get that back when we do the following hypofractionated regime, and need to take that into account. So we do it this way. You see 23Gy, 5Gy, so a big difference in this case. But when you look, you can see that we nicely can do this because we give the 23Gy really to this lesion in a nicely way, good conformity index in this situation.
And now this is the 5Gy isodose, the purple one. You can see, this is the cavity that we really nicely can shape the dose around the cavities. We can really give the 5Gy isotopes or give the 5Gy dose to the cavity, and give the 23Gy or 20Gy to everything else. These are the DVHs. You can see here, the 5Gy prescribed isodose or the cavity that gets 5Gy covering, the whole cavity. You know, the other lesions, the 20Gy, the 23Gy, and the overall conformity index, in this case, was 1.4. But to be honest, I excluded the cavity out of this calculation because first of all, it's a really large area, and this is not a typical situation. Conformity in this calculation doesn't make too much sense in this case, because this is nothing you will look at.
The V12Gy is large just because of this large volume of the cavity. And even the V3Gy situation is large. As I said, it's some kind of hybrid plan. The hypofractionated parts or the last six fractions for the cavity then in our department is done using the cranial SRS. The good point is we have already all contours within the elements. We can directly start and use all these contours within the same software. We have the dose there and everything, and we can take everything into account. You can see we prescribe them in this case, 30Gy because 5Gy were already given by the SRS session in this case. We did a plan, quite simple plan with only one table angle, but roughly complete arc, just not to influence too much what we already did for the SRS with the arc arrangements.
We try to be as simple as there, and not to have too much influence at that point. You can see here the DVH for the lesion at that point. And you can see down here in the cockpit, you can see that we used the old PTVs or the PTVs treated within the SRS session as organs at risk. In this case, we prescribed a strict maximum of 1Gy per lesion just to be sure to keep out all the dose out of the lesions, and not to have too much dose edit just by the hypofractionation, because we do not want to do this. If you look at the images down here, you can see the dose distributions here. You can see one of the lesions and there's one of the other lesions, and you can see that we even nicely shaped the 1Gy, which is the cutoff right now here, 1Gy around the old lesions, and we were able to spare them. So first, we can use the cavity already in the SRS session, and then we can use all the other lesions within the hypofractionated session as an alternative risk. This is how we do it in our department, and this is also something that is quite common right now.
Follow-up was, again, in November, and disease is controlled at that point. Patient is doing good and the disease is controlled. As we already did something around 100 patients, we try to think about what can we do with all this data. So we some statistics, some evaluation. At least we need to talk about risk of radionecrosis in this case, because everyone is aware of radionecrosis. We as well. So we look at all these things. But what I need to point out upfront is we do not see any increase in radionecrosis right now with our 100 patients, more than 100 patients so far. There's a 5% risk for all of these patients treated with SRS more or less. And we do see this, so their radionecrosis up to the progress, but we do not see an increase or so. That's just upfront, first of all. But when we want to talk about that, there's a nice paper out there already 2010, there are a lot of papers, but I really like this one because they did a nice study, and they also were looking at brain metastases in this case, and not on AVMs or so, so really on metastases in this case.
They were looking at radionecrosis and thought about irradiated volume as a predictor. As I said before V12Gy or V10Gy might be good predictor at that point. They really figured out that these might be really the predictors we should look at. They found out that V10Gy should be below 10.5 CC, or V12Gy if you want to look at that should be below 8 CC. If you get above that, you should think about hypofractionated treatment to lessen the risk of symptomatic radionecrosis. But as you've seen already, on my case reports, we do have patients with a V12Gy or a V10Gy that is higher than this strict number that they gave out in their paper. But we do not see an increase in radionecrosis.
That's why we thought about, radionecrosis is more or less a localized effect. It's happening surrounding the lesion you are treating. In the past, they were looking at patients that have one or two lesions, maybe three in maximum, but now we are talking about something else. So maybe we should not just look at these numbers, we should more think about having other numbers, maybe volume target, volume or something like that, or we should look at a localized V12 or V10Gy surrounding each lesion. Maybe we should end up with 14 V12Gy volumes for the lesions we treated or so. In our department, we just started plotting the V10Gy volume here for the whole-brain versus the total tumor volume to get an idea if there's some kind of correspondence or so.
Right now we found that there's a nice linear thing that you can bring in there. Right now we are thinking about, maybe we should more look at the total tumor volume and look at this, and if we are here on this and not up above this or below this, so if we are below this or on this curve, everything's fine. If we are above that, maybe we should reconsider something like that. Right now, we are just writing down V10Gy or V12Gy. We are not taking this as a strict criteria to do treatment this way or not, because we do not see any increase in that. So we're still in the process of evaluating everything, but this is already a nice start, I think.
Another thing that we are interested in is to have an idea of what about the gradient index? How good can we be, or how good should we be, at least. We also plotted the gradient index in comparison to the individual tumor volume in this case, and we figured out that most of our gradient index are between 5 and 3 for small lesions. You see if the lesion increases, the gradient index gets better and better and better. But this is just for us to have an idea, how good can we be, and if we got the plan, and it came out that we got a small lesion and or we've got a lesion of 3 CC, and we ended up with a gradient of index of 6, that we would take this plan and try to reoptimize it. Use another template, think about something else to be sure because we see we can do better at all.
Last but not least, we compared our gradient index so far and the conformity index so far that we can achieve to literature data. There's a nice paper published this year of this group here. They were also using the brain metastases element, and they were also doing VMAT plans, and did a comparison of that. I think they did the comparison of 8 plants or 10 templates, I'm not sure right now. We were in this magnitude, so we were able to achieve gradient index in the mean of 4.35, and a conformity index of 1.5. In comparison to them, we're on the same magnitude, we are even better, but I mean, our numbers are a lot larger. So if they would do it for more numbers, I think their statistics would be very, and maybe they would be in the same magnitude.
We try to be as good as this at least. In the end, to come up to a conclusion, what I wanted to point out here as our experiences that treatment of multiple brain metastases with only one isocenter is a challenging approach even for planning and delivery, as we have seen. You have to deal with this. But the good point is the Brainlab software can handle the planning process, and offering a very fast and reliable solution. The next point is due to the automatic process, and the limiting or planning variabilities in this situation. It's nice that you can really reduce the planning time and you would create very consistent, but still individualized plans for every patient. The ExacTrac system is able to compensate all possible misalignments. That's what I hopefully have shown you. The take-home message is with both systems together, patients with multiple brain metastases can easily be treated with SRS and one isocenter in Linac-based approach. So thanks for your attention. As you already have asked quite some questions within the chat, I will just briefly go through that, and try if I can answer all of these.
So the first question was, how long did it take you to commission the multiple brain metastases SRS element, and what steps did you perform? The first nice point about this is if you want to get the system running, and you're already iPlan customer, the good point is that all the data you already took and got will be the same. You will end up with the same machine profile that you already created for the iPlan. If you did all this, you should take into account for sure that you have to deal with small field sizes. You have to be sure that you have the right detector, maybe a diode, unshielded diode for the small lesions or diamond detector and so on. If you did all this, then the next step would be within the commissioning process to to see if your software behaves exactly the way that you would like to. The first point is there's an element where you can easily do the commissioning. So you can check your system by using the element part of that, by creating beams or even arcs, map them on the phantom, and then easily export them and measure them if you want to.
In our department as this element was still in development when we started with this, we created a couple of library plans that we then with single lesions, with multiple lesions, and we put them on...we mapped them on a phantom, and then measured them and recalculated them as well in Eclipse, and did the commissioning process more or less in this way. Another point is that part of the commissioning, at least if you're handling the Monte Carlo, is already done by Brainlab. This is also quite nice because you have to send all your data to them, and they will prepare the files for you for the Monte Carlo approach, and they will already end up with nice PDF document that includes a lot of statistics profiles and so on. Most of them is done by using library plans, and doing different treatment plans.
The next question is what kind of optimization parameters or structures have you used for the Eclipse plans in your plan comparison? This is a good point. I'm pretty sure that there are a lot of people out there that will be better maybe doing the treatment first, because this was already done in the past. We did it in 2014 or so, and maybe we were not that good in this case. We used ring structures surrounding the volumes to be sure to lower the dose at that point, and to use this during the planning optimization process to keep the dose conformity nicely. That's why we achieved a good conformity index, but the normal tissue sparing at that point was not that good. So we used mostly the ring structures, and then introduced some more helping structures to, which we brought in between two lesions, so they're two lesions within the same plane. I introduced another helping structure between these two lesions to try to lower the dose, the low-dose shower at that point.
It was a lot of work, but it didn't quite nicely worked out in the end, but, yeah. This is more or less what we did. The good point is, in the Brainlab solution, you do not have to think about this and deal with this. This will all be done automatically in the background. The next question would be, how long did it usually take you to generate an iPlan plan with individual isocenters? I think for this case I presented to you, for the case with the six lesions, it took me roughly two and a half to three hours. For 10 lesions, it easily increases to a full day because of all the problems that adding another isocenter influences the dose to every other isocenter. I wanted to be as strict and as accurate as possible to really only give the 20Gy I prescribed and not 22, or just because of different influences. As I had to deal with all table angles on my own, it roughly took me two and a half hours. But for the situations with 10 lesions, I think I ended up with seven hours or so. Not sure, but it took really quite long time.
When you think about the idea that you will end up with the same time on the machine, because you have to treat every single isocenter on its own, that might not be the best approach. How often do you image the ExacTrac for mets cases? What is your IGT workflow for understanding intra-fraction motions? Do you use Cone Beam CT for setting up your cranial patients too? No, we do not do Cone Beam CT. We only use ExacTrac, and ExacTrac is used in our department for all head cases. So not only the brain metastases cases, also, doesn't matter if it's a glioblastoma or whatever, we use ExacTrac for all of these patients. We do upfront imaging, positioning correction. After moving the table, we do another imaging to be sure that we applied the shifts correctly, and that the machine applied the shift correctly, and that everything is okay. At that point, we do another check, and see we are within the tolerances 0.6 millimeters and 0.6 degrees of rotation in our situation for SRS cases. The workflow is at that point, we do... Whenever we change the table angle, whenever we add a non-coplanar, move to a non-coplanar arc, we do imaging again when we reach the table position.
We do imaging, we do ExacTrac correction, we apply the shifts, we do imaging again, to be sure that shifts are applied correctly, and to be sure that we are in position again, and then we irradiate that. This is done for every single arc and every single table position. We do it for every table position. Well, image-guided radiation workflow for understanding infra-fraction motion. It's more or less like this. Whenever there's something, we would see that, but we do not see any real. Within the brain or within the skull, we do not see any motion. There shouldn't be any motion at all for brain metastases. So we do not see anything like that. But as I said already, we do imaging upfront before we treat every single table angle or every single arc to be sure that still everything is in place because we are using a mask system, and there might be room for the patient to move within the mask.
What clinical margins do you place for PTV definition for multiple brain metastases cases? And, is this a uniform CTV expansion? What margins have you used for iPlan? In the past we used margins for iPlan of 1 or 2 millimeters depending on how good the lesion was visual. If it's a cystic lesion, and the borders of the lesion cannot be clearly defined or so, we end up with a 2-millimeter margin, same for elements as well. But normally we only apply 1-millimeter uniform margin most of the time. If we are pretty sure, imaging is really nice and new, and MRI scans were done just a day before treatment starts or whatever. We also end up in some situations with a margin of zero depending on how good we can really visualize and define the lesion. Margins are only there for compensating for problems during delineation of the lesion, not for the table or so on, because we are within the 0.6 millimeters for all of that.
There's another question. How do you verify the rotational accuracy? And then in particular the rotation of calibration of ExacTrac. I'm not quite sure what you mean by rotational calibration of the ExacTrac. But we do verify the ExacTrac by using the software or the in-house or the tools that were offered already by Brainlab. There are a couple of phantoms and a couple of tools that has to be used. You have to check the radiation isocenter or the calibration of the radiation isocenter on a regular basis for the software. We do Winston-Lutz test upfront. This has done or for all cases upfront. And so we do this full-day end-to-end test on a regular basis that really includes also positioning of the patient using the ExacTraC in the full workflow.
So we have the BB, and then we have the phantom there with an iron chamber and also with films. And then we irradiate that, and see if everything's in place. We do not do that once per week or so, but we do it on a regular basis at least every three month to check that everything is okay at that point. Hopefully, I answered your question at least a little bit.The next question would be for patient-specific quality assurance, what is your passing rate when comparing the monitor units with either Eclipse or your in-house software? The passing rate for portal dose is two and two, in this case, we use two and two. The passing rate for comparing the monitor units with our in-house software is also 2% at that point.
For Eclipse, it's a little bit more complicated because the situation in our institution is that our Eclipse is not commissioned for a pencil beam. We did AAA and Acuros as there is another question down below here in the questions. There might be already a difference just because of a different dose algorithm. At that point we say, okay, for lesions that are more or less central located within the brain, we are within 2%. But for lesions that are more next to the skull or so that are more next to the inhomogeneities, we can also end up with 3% or 5%. The recalculation is just to see if there's really something very, very strange. But most of the time we end up with roughly in the magnitude of 3%, even if we use a AAA or not pencil beam at that point.
Another question was, for your multiple mets plans and elements, how many of the plans required repeating gantry arcs to cover all lesions? Also if gantry arcs have to be repeated, how could the delivery time be comparable to VMAT RapidArc? The delivery time... It depends a bit. So I've no number really for how many of the plans require repeating gantry angles. Repeating gantry angles is most of the time more effect of giving enough dose to some of the lesions, as you said, in your question, or if you have a lot of lesions that are all within the same plane, you were doing a clockwise rotation, and you already doing three of the lesions, and doing a counterclockwise rotation, and irradiating another three and so or two lesions and not the same, maybe. One of the processes within the optimization process is to cover as much lesions as possible with as many RX as possible to streamline the process.
The amount or the treatment time can still be comparable to VMAT due to the fact for the RapidArc solution, you do a volumated...so an IMT approach. A lot of your monitor units were just going within the MLC or within your drawers without even adding any dose to the patient. When you do VMAT, your dose will be... At some point, the dose rate will go down from maybe if you have 600 monitor units as default, will go to 300 or something like that, while doing dynamic conformal RX, your dose rate will be stable at 600 monitor units. You will still always give 600 monitor units. In the end, dose and the patient is directly dependent to monitor units. You end up with the same amount of monitoring units or roughly the same, so you end up with the same at least in the same magnitude of delivery time.
The next question was something that I already answered a little bit maybe, but I would just go to that. Both AAA and Acruos were shown to have issues for calculation when targets are small, roughly 1 cm. There's the paper for [inaudible 01:16:15] medical physics, 2011. Do you see systematic differences from elements multiple brain metastases when you do your check in Eclipse? What are your smallest fields measured for Eclipse? As I said already there are differences. This might be also depending on that effect, but our...the good point is at the same time when we did the commissioning for the iPlan and the elements or iPlan, to be honest, because we started with iPlan, we also did the commissioning for the Eclipse system, because it also came new into our department at that point, and we used the same more or less. We included measurements in the Eclipse down to 1 centimeter by square field size for our protectors and so on.
We measured that to quite small field sizes, not just down to 3 by 3 or so, we went down to 1 by 1 as we did the measurements already for the iPlan, and they were there. But in the iPlan, we go down to 0.5 by 0.5. These are the smallest fields that we had in the Eclipse. That's why our results might be quite nicely, maybe. Why do you monitor 3Gy for your patients in addition to 12Gy and 10Gy? We do this just to have an idea what the low dose is given to the patient, just to have an idea how much low dose we were giving in terms of... This is a little bit like a quality parameter for all our treatment planning.
We do not want to give too much low dose to everything else. That's why we monitor. There's no real clinical reason right now, or something like that. This is just as a parameter to monitor, to have an idea, how much low dose, what about the low dose spillage that we were giving to the patient? In addition to the other ones that we still think about radionecrosis also, this has no directly link to clinical data so far, but it's a parameter for us just to have an idea of plan quality. Because we want to keep this as low as possible because there do not need to be so much dose in the normal brain.
Next question would be, how many templates do you use clinical, and how do they differ? That's a really good question, because I think when we started with the prototype, we started with roughly 30 templates or something like that, because upfront, we thought about all situations, and we created templates for almost every situation, and about everything that we would think would come into our department. Right now I think, we ended up with eight or nine. We got templates for situations where we would only use one template. 5RX equally split more or less around the skull. And then we have templates for situations where we only would like to treat more on the left or more on the right side with RX more on that side. And then we have templates that have different numbers of minimal used arcs. One point within the templates, or one point within the optimization process is that the software looks at the numbers of arcs used, or that needs to be used.
We found out that we had situations where the system came up with a solution that three arcs would be sufficient to give the dose to the patient. You end up with a nice conformer plan, but we thought about maybe it would be better to end up with 4RX or even 5RX, and then we have different situations where we forced the system to use at least 4RX or 5RX. This makes another... So you have to have every other templates multiplied by this, so we end up with nine or so. But we mostly use the 5-arc template, and give the software as much freedom as possible to come up with 3RX, if necessary, and if that's enough. It's pretty easy. You do not need to have