- 435
Transcript
So thank you, Giuseppe, for the nice introduction. Yeah, as we do not have so much time, I will start directly. So I want to talk about intra-fraction motion for cranial radiosurgery of patients. As the topic of the symposia is "How does every single millimeter count," so the first millimeter or the millimeter at the machine might be, maybe the intra-fraction motion or so. And why does it count or why is it important? I, hopefully, will give you an idea about that within the next 15 minutes. So here's my little disclosure.
So one of the elements that are out on the market is the multiple brain metastasis element. And, in this case, we are dealing with multiple targets, and we will do a treatment plan and cover them with a single virtual isocenter. So, not like in the past where each lesion got its own isocenter, now we got a virtual isocenter, and the isocenter will hardly be within one of the lesions. So if there might be intra-fractional motion now, this can lead to large deviations in the dose distribution because every rotation or every millimeter that is a shift will result in even a larger effect with increasing distance of the lesions to isocenter, as you can see down here.
So, what we did then in our institution to get an idea how large are these shifts, and what might be important about these is that we started to do an analysis of shifts for patients treated with SRS in our department on a Novalis STx during 2017 and 2018. And we evaluated the correction values that we got from the ExacTrac System in terms of number of necessary corrections that we had to do, in terms of the correction values in 6D. And then after doing some analysis of that, we did recalculations of treated patients with these values derived from that to get an idea what is the influence on the dose distribution on the deviations and so on because that's what we, in the end, are interested. Do we have to really take this into account, do we have to correct for this and so on?
This is just a little part of the file we derived from that just to give you an idea how this looks like. Here are the different shifts. We looked at the couch angle, the planned couch angle at the couch kick and in terms of change in the couch position. And you see already everything that is red was not within our tolerances so we had to do a repositioning, a check of that imaging, use the 6D robotics couch. And here are just four patients randomly chosen from our cohort, and you see, we had to do corrections in almost every cases despite this patient where everything was from scratch already within our tolerances, just to give you an idea how that is spread around that. But we will come to that in a couple of seconds later.
So what did we analyze? We use in our department for the ExacTrac limits, so for SRS, 0.5 millimeters for translational corrections or for translational resolution, and 0.5 degrees for rotation in terms of pitch, roll, and yaw. And in the table next here on the left, you can see we had 125 patients that we evaluated with the number of roughly 400 couch kicks in total. And the first really interesting number might be here, because we had to do for really almost 66% of all couch kicks, we had to do after we applied the couch rotation, we had to do a recheck, had to do a repositioning and so on to be really within our limits that we had here.
Here are the values given in mean before correction and after correction, and you see using the corrections we can really easily even half the discrepancy in these terms. So to give you an idea about the spread of these situations, we have here the number of treated patients. So, up to 125 in this situation, and the mean 3D deviation in red that we had after couch, that we had without the correction, and blue then for the situation with corrections. And you see that we easily can reduce the correction values or can use the position and the accuracy in the mean really to be below 0.5 so to be really within stereotactic circumstances in our situation.
You see there are easily a lot of patients that are really with very large deviations here in the 3D mean and there are some that are quite nicely positioned here. If we split that up into vertical, longitudinal, and lateral deviation, to give you an idea about the distribution of that, you can see easily that after repositioning, we are really able to bring everything within a nice position, with the maximum number really being at 0 or more or less 0 millimeters difference, and the same for the rotational errors here for yaw, roll, and pitch deviation.
Just to give you an idea how that might look like in 3D and to give you...I think, because this really nicely visualize the spread of the values that we got, so this is before we did the repositioning and correction for vertical, lateral, and longitudinal, and then we did the repositioning, and you see that really nicely everything is spread around in a nice manner around 0 millimeters with a maximum of 0.5 millimeters, which is within our tolerance.
So the question really behind this is why do we have these differences? What is the reason for this? Is it maybe a systematic problem? Is our couch not okay? Do we have some problems with the couch, with the rotation of the couch, because everything like this happens after couch kicks so after rotational things, or is it the patient, or whatever? And to get an idea about that, we did extensive checks for the couch so we did a lot of Winston-Lutz test where we checked...changed just the couch rotation, where we added also gantry rotations, collimator rotations to get some kind of a real treatment patient situation. And the answer to this is there's no significant increase in misalignment when just adding the couch rotation as well.
So our couch really rotates within 0.2 millimeters if we do a full rotation of the couch. And it doesn't matter if we do it once per week or even once per month, the couch is really stable from day one from installation up to now. So the couch is not the problem. The reason for this might be the patient, because that's the only thing. So we use a mask system in this case, a stereotactic mask, and the patient can move within the mask for sure. Okay? And the patient can move within the mask for sure. Even if the patient doesn't move within the mask, so we want to be below 0.5 millimeters. Even a little shift within the shoulder region or whatever when we turn the couch, can also really easily already start to introduce some kind of error.
So how do these shifts then influence the dose distribution that we see here in this case. So to keep in mind, whenever there's something red, we had to do a recalculation and a recheck for that. So what we did then is we did a recalculation of one of our patients, in this case, Patient 4 out of our statistics, and we just introduced really the shifts, the mean shift that we derived from our values, in this case, on the isocenter, and shifted the patient in that way. We even introduced the yaw error, so the couch rotational error in this situation, and did a recalculation. And you see even with adding the mean shift, which are 0.5 millimeters or even below 0.5 millimeters, which is within our tolerances or would be in our tolerances in total, you see already a difference in the dose distribution for the lesions when you look at the deviates. Here is just a little image of the dose distribution for the patient themselves for one of the lesions and to see easily that the dose maximum is a little bit higher that the dose distribution shift out of the lesion, so you would give a really prescribed dose to the normal tissue, so to the healthy brain, and will under-dosage your lesion.
But this is for the mean, and our idea is, okay, then let's see, if we do not have any mean value or so, what would happen if we add the real shifts that we saw during this and during the treatment of the patient. And now we added the real shifts for the patient given down below here, and you see we got a lateral shift of 2.5 millimeters, in this case, for the first couch kick, which is quite high, and then you see how much...well, how often we did new images before we...until we get to a nice position there.
And this is the dose distribution in terms of the deviate for adding the real shift that we really saw for this patient, just to give you an idea what would happen if we treated this patient, exactly this patient without any correction. And if you easily can see when you look at the deviates, so the lines with the square signs are the original ones or the ones shifted just by the normal values, so the original ones, and the other ones are the values that we derived from this really...recalculation. And, easily, you can see within the deviate that you really got large under dosages already for your lesions.
But let's look at the dose distribution itself. On the right...on left, again, the original one, and then on the left, the shifted one, and you see that you easily just irradiate half of the lesion in this case here with the prescribed dose and everything else will get less dose than you wanted to achieve. And the rest of the maximum, because it's just...the shift really gets nicely outside of the lesions so what will be the end? If we wouldn't do any correction in this case, we will under dose the lesions so we will get a nice recurrence at the position of the lesion, a nice relapse there.
And if we get really in trouble, we will get a nice radionecrosis next to the situation where we got the prescribed dose to the healthy tissue. So that's a situation we do not want to deal with afterwards. For the other lesion, just to give you an idea, this lesion is even farther away from the isocenter. The effect will get worse and worse, because it's just a geometrical thing.
So the next thing that we then thought is, okay, this is our institution, these are our results, what about other institutions? Might this just be a problem of our cohort? Are we not good? Is there any shift...or our mask is not good, or whatever, or treatment with the patient in the wrong way. So we did a comparison with published data out there, and we compared to three papers at least that we're looking at more or less the same effects. So they evaluated in phantom studies, not with real patient data, but in phantom studies, they evaluated, in some situations, even the lateral, longitudinal, and vertical shifts, and we see we are in more or less in the same dimension like that. So it's not our problem, it's a general problem so that there is intra-fractional motion and we do have to take that into account maybe.
So as we know there are shifts, how can we deal with these? I mean, the first and most simple idea would be, "Okay, we just add margins." So there's a shift, let's add a margin until our lesion is large enough or our PTV is large enough to really be covered, it doesn't matter if there's a shift or not. Or, do we need something else, image guidance and repositioning? And I'm pretty sure Giuseppe will talk about margins quite a lot. But just to give you an idea that adding margins might be not the real approach, here's a little study that we did for a patient with five lesions. And we added 0 millimeter margin to the GTV, 1 millimeter, 2 millimeters, 3 millimeters to the lesion.
And here, down here, you see a dose distribution of that. And we looked at the total PTV volume and the V12 for the normal brain, because that's more or less an indicator for radionecrosis, or the risk of radionecrosis.
And I think if you look at the numbers of the increase by just adding a margin of 1 millimeter or 2 millimeters or 3 millimeters, it's pretty obvious that adding a margin cannot be the solution to this problem because adding a margin means irradiating more tissue and irradiating more healthy tissue because the GTV, our real tumor, will stay the same. We will just add a margin to create a bit larger PTV. So adding a margin is not the right idea. We need to do something else.
And I would really strongly suggest that we need to do image guidance in this situation. So do imaging, look at the shifts, do repositioning using 6D robotic couch, whatever. Do imaging again to ensure that you're really treating the right lesion, the right area, and then treat the patient. And what might be out there to use, what other technical solutions for image guidance. So, more or less, we got the OBI. In our institution, we got onboard imaging, and we got the ExacTrac system. And the good point is the ExacTrac system is room based so we can use it whenever we want.
We can do...use it for every couch angle despite the fact the OBI is not usable at every couch angle because whenever you use non-coplanar arcs, you will get into trouble. ExacTrac is quite fast. You can do repositioning for everything, and you can really use the 4Pi full hemisphere for imaging and everything like that.
So, to conclude, treatment of complex multifocal lesions is challenging, even the treatment planning, but also, on the machine, when we think every hard work is done already, we have to look at that as well. It really creates a high demand on the positioning accuracy due to this virtual isocenter. So with increasing distance of the isocenter shift, the effect of shifts will get harder and harder. And if we really want to compensate for that...for intra-fraction motion, the only way it can be, to be using imaging and repositioning for every couch kick. So what I really strongly would advise you is do imaging, do repositioning, do, again, imaging to be sure that repositioning was in a good way.
As you had saw, maybe, from our table there, there's a couple of situations where we had to do multiple imaging for the same couch angle, because there were still shifts so we have to do that afterwards to be sure that everything is in a good position. And I would strongly advise to use some kind of tool like the ExacTrac System, because that can handle this really independently of the couch or gantry angle.
So, thanks for your attention.
So one of the elements that are out on the market is the multiple brain metastasis element. And, in this case, we are dealing with multiple targets, and we will do a treatment plan and cover them with a single virtual isocenter. So, not like in the past where each lesion got its own isocenter, now we got a virtual isocenter, and the isocenter will hardly be within one of the lesions. So if there might be intra-fractional motion now, this can lead to large deviations in the dose distribution because every rotation or every millimeter that is a shift will result in even a larger effect with increasing distance of the lesions to isocenter, as you can see down here.
So, what we did then in our institution to get an idea how large are these shifts, and what might be important about these is that we started to do an analysis of shifts for patients treated with SRS in our department on a Novalis STx during 2017 and 2018. And we evaluated the correction values that we got from the ExacTrac System in terms of number of necessary corrections that we had to do, in terms of the correction values in 6D. And then after doing some analysis of that, we did recalculations of treated patients with these values derived from that to get an idea what is the influence on the dose distribution on the deviations and so on because that's what we, in the end, are interested. Do we have to really take this into account, do we have to correct for this and so on?
This is just a little part of the file we derived from that just to give you an idea how this looks like. Here are the different shifts. We looked at the couch angle, the planned couch angle at the couch kick and in terms of change in the couch position. And you see already everything that is red was not within our tolerances so we had to do a repositioning, a check of that imaging, use the 6D robotics couch. And here are just four patients randomly chosen from our cohort, and you see, we had to do corrections in almost every cases despite this patient where everything was from scratch already within our tolerances, just to give you an idea how that is spread around that. But we will come to that in a couple of seconds later.
So what did we analyze? We use in our department for the ExacTrac limits, so for SRS, 0.5 millimeters for translational corrections or for translational resolution, and 0.5 degrees for rotation in terms of pitch, roll, and yaw. And in the table next here on the left, you can see we had 125 patients that we evaluated with the number of roughly 400 couch kicks in total. And the first really interesting number might be here, because we had to do for really almost 66% of all couch kicks, we had to do after we applied the couch rotation, we had to do a recheck, had to do a repositioning and so on to be really within our limits that we had here.
Here are the values given in mean before correction and after correction, and you see using the corrections we can really easily even half the discrepancy in these terms. So to give you an idea about the spread of these situations, we have here the number of treated patients. So, up to 125 in this situation, and the mean 3D deviation in red that we had after couch, that we had without the correction, and blue then for the situation with corrections. And you see that we easily can reduce the correction values or can use the position and the accuracy in the mean really to be below 0.5 so to be really within stereotactic circumstances in our situation.
You see there are easily a lot of patients that are really with very large deviations here in the 3D mean and there are some that are quite nicely positioned here. If we split that up into vertical, longitudinal, and lateral deviation, to give you an idea about the distribution of that, you can see easily that after repositioning, we are really able to bring everything within a nice position, with the maximum number really being at 0 or more or less 0 millimeters difference, and the same for the rotational errors here for yaw, roll, and pitch deviation.
Just to give you an idea how that might look like in 3D and to give you...I think, because this really nicely visualize the spread of the values that we got, so this is before we did the repositioning and correction for vertical, lateral, and longitudinal, and then we did the repositioning, and you see that really nicely everything is spread around in a nice manner around 0 millimeters with a maximum of 0.5 millimeters, which is within our tolerance.
So the question really behind this is why do we have these differences? What is the reason for this? Is it maybe a systematic problem? Is our couch not okay? Do we have some problems with the couch, with the rotation of the couch, because everything like this happens after couch kicks so after rotational things, or is it the patient, or whatever? And to get an idea about that, we did extensive checks for the couch so we did a lot of Winston-Lutz test where we checked...changed just the couch rotation, where we added also gantry rotations, collimator rotations to get some kind of a real treatment patient situation. And the answer to this is there's no significant increase in misalignment when just adding the couch rotation as well.
So our couch really rotates within 0.2 millimeters if we do a full rotation of the couch. And it doesn't matter if we do it once per week or even once per month, the couch is really stable from day one from installation up to now. So the couch is not the problem. The reason for this might be the patient, because that's the only thing. So we use a mask system in this case, a stereotactic mask, and the patient can move within the mask for sure. Okay? And the patient can move within the mask for sure. Even if the patient doesn't move within the mask, so we want to be below 0.5 millimeters. Even a little shift within the shoulder region or whatever when we turn the couch, can also really easily already start to introduce some kind of error.
So how do these shifts then influence the dose distribution that we see here in this case. So to keep in mind, whenever there's something red, we had to do a recalculation and a recheck for that. So what we did then is we did a recalculation of one of our patients, in this case, Patient 4 out of our statistics, and we just introduced really the shifts, the mean shift that we derived from our values, in this case, on the isocenter, and shifted the patient in that way. We even introduced the yaw error, so the couch rotational error in this situation, and did a recalculation. And you see even with adding the mean shift, which are 0.5 millimeters or even below 0.5 millimeters, which is within our tolerances or would be in our tolerances in total, you see already a difference in the dose distribution for the lesions when you look at the deviates. Here is just a little image of the dose distribution for the patient themselves for one of the lesions and to see easily that the dose maximum is a little bit higher that the dose distribution shift out of the lesion, so you would give a really prescribed dose to the normal tissue, so to the healthy brain, and will under-dosage your lesion.
But this is for the mean, and our idea is, okay, then let's see, if we do not have any mean value or so, what would happen if we add the real shifts that we saw during this and during the treatment of the patient. And now we added the real shifts for the patient given down below here, and you see we got a lateral shift of 2.5 millimeters, in this case, for the first couch kick, which is quite high, and then you see how much...well, how often we did new images before we...until we get to a nice position there.
And this is the dose distribution in terms of the deviate for adding the real shift that we really saw for this patient, just to give you an idea what would happen if we treated this patient, exactly this patient without any correction. And if you easily can see when you look at the deviates, so the lines with the square signs are the original ones or the ones shifted just by the normal values, so the original ones, and the other ones are the values that we derived from this really...recalculation. And, easily, you can see within the deviate that you really got large under dosages already for your lesions.
But let's look at the dose distribution itself. On the right...on left, again, the original one, and then on the left, the shifted one, and you see that you easily just irradiate half of the lesion in this case here with the prescribed dose and everything else will get less dose than you wanted to achieve. And the rest of the maximum, because it's just...the shift really gets nicely outside of the lesions so what will be the end? If we wouldn't do any correction in this case, we will under dose the lesions so we will get a nice recurrence at the position of the lesion, a nice relapse there.
And if we get really in trouble, we will get a nice radionecrosis next to the situation where we got the prescribed dose to the healthy tissue. So that's a situation we do not want to deal with afterwards. For the other lesion, just to give you an idea, this lesion is even farther away from the isocenter. The effect will get worse and worse, because it's just a geometrical thing.
So the next thing that we then thought is, okay, this is our institution, these are our results, what about other institutions? Might this just be a problem of our cohort? Are we not good? Is there any shift...or our mask is not good, or whatever, or treatment with the patient in the wrong way. So we did a comparison with published data out there, and we compared to three papers at least that we're looking at more or less the same effects. So they evaluated in phantom studies, not with real patient data, but in phantom studies, they evaluated, in some situations, even the lateral, longitudinal, and vertical shifts, and we see we are in more or less in the same dimension like that. So it's not our problem, it's a general problem so that there is intra-fractional motion and we do have to take that into account maybe.
So as we know there are shifts, how can we deal with these? I mean, the first and most simple idea would be, "Okay, we just add margins." So there's a shift, let's add a margin until our lesion is large enough or our PTV is large enough to really be covered, it doesn't matter if there's a shift or not. Or, do we need something else, image guidance and repositioning? And I'm pretty sure Giuseppe will talk about margins quite a lot. But just to give you an idea that adding margins might be not the real approach, here's a little study that we did for a patient with five lesions. And we added 0 millimeter margin to the GTV, 1 millimeter, 2 millimeters, 3 millimeters to the lesion.
And here, down here, you see a dose distribution of that. And we looked at the total PTV volume and the V12 for the normal brain, because that's more or less an indicator for radionecrosis, or the risk of radionecrosis.
And I think if you look at the numbers of the increase by just adding a margin of 1 millimeter or 2 millimeters or 3 millimeters, it's pretty obvious that adding a margin cannot be the solution to this problem because adding a margin means irradiating more tissue and irradiating more healthy tissue because the GTV, our real tumor, will stay the same. We will just add a margin to create a bit larger PTV. So adding a margin is not the right idea. We need to do something else.
And I would really strongly suggest that we need to do image guidance in this situation. So do imaging, look at the shifts, do repositioning using 6D robotic couch, whatever. Do imaging again to ensure that you're really treating the right lesion, the right area, and then treat the patient. And what might be out there to use, what other technical solutions for image guidance. So, more or less, we got the OBI. In our institution, we got onboard imaging, and we got the ExacTrac system. And the good point is the ExacTrac system is room based so we can use it whenever we want.
We can do...use it for every couch angle despite the fact the OBI is not usable at every couch angle because whenever you use non-coplanar arcs, you will get into trouble. ExacTrac is quite fast. You can do repositioning for everything, and you can really use the 4Pi full hemisphere for imaging and everything like that.
So, to conclude, treatment of complex multifocal lesions is challenging, even the treatment planning, but also, on the machine, when we think every hard work is done already, we have to look at that as well. It really creates a high demand on the positioning accuracy due to this virtual isocenter. So with increasing distance of the isocenter shift, the effect of shifts will get harder and harder. And if we really want to compensate for that...for intra-fraction motion, the only way it can be, to be using imaging and repositioning for every couch kick. So what I really strongly would advise you is do imaging, do repositioning, do, again, imaging to be sure that repositioning was in a good way.
As you had saw, maybe, from our table there, there's a couple of situations where we had to do multiple imaging for the same couch angle, because there were still shifts so we have to do that afterwards to be sure that everything is in a good position. And I would strongly advise to use some kind of tool like the ExacTrac System, because that can handle this really independently of the couch or gantry angle.
So, thanks for your attention.