- 718
Transcript
Thank you. First, I'd like to start by thanking Brainlab and Varian for the opportunity to share our experience that we've achieved with the HD-MLC. This is the gang I've been working with, our radiation oncologist, our neurosurgeon, and my fellow physicist if you haven't figured that out already. We're basically two linac standard RT operation that's upgrading to radiosurgery next fall. And last fall, we had our Trilogy TX installed. And then, later on in January, we had our HD-MLC installed. And that's when our Trilogy TX magically turned into our Novalis TX. And we've since then treated 90 patients, 30% of which are IMRT. And we've had no basic issues, no leaf motor changes. It has been a very stable system. And only on three occasions have we had to switch patients because of field size limitation to our...switch patients to our other linac. And here's the token small leaf picture.
And you probably have seen a lot of these and this as well. Again, the HD-MLC is a lot less rounded than the Millennium MLC and has about 5% more tungsten in it. And we've found that the motor stability on the MLC has been very good. We've had no leaf replacements in six months, and we expect that if we do have one, it'll take a lot longer than a typical Millennium motor since these motors are so much smaller. And we expect them to last longer because the motors are new and improved. The leaf speed doesn't travel at maximum speed at all times, and we don't expect leaf collisions. We didn't have them before at the Millennium, but they have more collision prevention tests during the MLC initialization.
And we've found that with some geometric field shape studies, as well as Picket fence studies and Dynalog studies that are statically position precision is about a 10th of a millimeter. I don't think Varian is quoting this, but that's what we've seen and that the dynamic leaf position precision is one millimeter, according to Dynalog results. Here's a slide on the transmission for the HD-MLC. We have a film on the left, on the right our profiles against and with the leaves. And you can see that the transition for the HD is about 50% lower than the Millennium, and the inner leaf leakage is about half that of the Millennium. And here are our actual transmission values for... And what we've put into Eclipse or configured it with is the 10 by 10 field size value for the transmission, which is about 1 for the 6 MV HD-MLC and about a 1.2 for the 10 MV HD-MLC.
And as you can see, there's a bit of a field size dependence there. Next, we measured the dosimetric leaf gap and got similar results to Fang-Fang. Basically, we use the sliding window technique where we varied the gap of the window and took our chamber measurements for each gap, and then extrapolated to a zero reading to get a dosimetric leaf gap size of 0.8 millimeters. I don't have the units here, but it's millimeters for the 6 MV and about 0.9 for the 10 MV. However, we took it one step further and decided to see if changing our DLG from the sliding window value would improve our dosimetric results so that if we measured the dose distribution for the field, we'd get a better comparison with the Eclipse calculated distribution.
And for that, we used a step pattern, which doesn't have a jagged distribution that might make Eclipse calculation resolution an issue. It doesn't have any tongue and groove issue, and it's not sensitive to positioning. So if we're changing our DLG, we're not compensating for any of these other factors. So basically, what we did, was reconfigure Eclipse with a different DLG value, re-optimized, recalculated the leaf motion. And one point I wanted to make is changing the DLG is not going to change your dose distribution. It's going to change your final leaf motion. Then we remeasured the dose distribution and compared the measured to calculated. And this is the system that we used. Basically, the MapCheck dose calibrated to give absolute dose, and it's at a depth of seven.
And here's our first result for the 6 MV HD. We have measured versus calculated 4.8 sliding window value to a 0.4. And we didn't see much difference. This is an error map on the right for both of them showing whatever is lower of the distance to agreement or the percent dose difference, and the mustard green and the light green here correspond to the 3-millimeter 3%, 3-millimeter distance to 3% dose difference values. So there's not much difference. However, for the 10 MV, when we went from 0.9 to 0.5, we had a marked improvement in our error map.
And we repeated this for a 10 MV for a typical prostate fluence for a single field. And again, you can see the improvement of the 0.5 value for the dosimetric leaf gap used in Eclipse versus the 0.9 we get from the sliding window. And so we did more studies showing the dose difference between measured and calculated for the different steps and for different DLG values and found that we got the best agreement for about the 0.5-millimeter value. And these are, again, our final DLG values. We're also very interested in the HD fluence delivery as well as dose verification. We were somewhat concerned that we might see more tongue and groove with the HD-MLC. So we took several transmission films. Here we have the HD on the top and 120 on the bottom. The tongue and groove here is represented by the white line and happens when one leaf is moving against a static one.
And as you can see, we don't have more white lines with the HD than the M-120. And if we take a profile here, we get an interesting plot where you see that the HD is a lot smoother. This was for a 45-degree angle, and because of the higher resolution is much smoother and has a similar tongue and groove right here, down here. So it does seem to perform much better at the fluence level. We also, for per patient QA, we measured the dose distribution for all the patient fields using the same method I mentioned before, and we've bend our results. So these are for all the fields that we've given our patients for the different systems and energies, the different MLC models and energies. And these are the percent dose difference NDTA we've been reporting to what passed, had 95% of the points pass.
So we see that most of the field dose distribution measurements are here within a 4% dose difference in two DTA, and it's very similar for all of them. So we see that for the HD, we're not doing any worse than we used to, which is really important when your practice is your regular radiation therapy. You want it to at least be the same as it used to be, if not better. So we're not getting anything out here at the higher percent dose difference or higher DTAs. So that was encouraging. Next, We took a look at a lot of dose distributions using the differently sizes. And a lot of people have done that before with different MLC models and different treatment sites. We decided to look at it again anyways and gotten similar results as well as slightly different results. And first, to start off, two sort of similar results. Using the 2.5 leaves versus the 5, we did a regular prostate without nodes, and the distributions are very similar as are the DVHs for the organs at risk. And then similarly, for a complex head and neck with dose painting, 95% around the high dose PTV, 83% around the lower dose PTV. We get similar distribution, a little less hot on smaller leaves and down here.
However, when we looked at our DVHs, we found that the 5-millimeter actually gave slightly better DVHs in this dose range for the brainstem and the cord. But then we got when we went to a much smaller concavities and higher dose gradients, that's when we saw the advantages of the 2.5 millimeter MLC, and our first case was an ethmoid sinus. And here's one of the field fluences. And you can see for the higher resolution HD, how in the small 1-centimeter type concavity, how much better it does than the other models and a block of the other models are, and that translates overall into these distributions. And we have a sharper gradient for the HD 2.5 millimeter. So the 70% doesn't go beyond these lines, whereas the others do. And also, the 10% is curving in here.
And so we're getting more lens sparing, and we see that in the DVHs. And so, in all cases, the brainstem and the lens, we have better sparing with the HD, and for the PTV, we have a more homogeneous dose distribution. And then we did one more. We haven't done any of these cases, but these are my test case, my test PTV. This patient is healthy. There's no tumor. So we did a situation on this side. It's the 5-millimeter. This is the 2.5, and for low cord constraint and low gradient, we get pretty similar results in terms of a hotspot and coverage and gradient. But this changes a lot once we start using a high dose constraint in both situations to end up with a 2.7-millimeter drop-off. And where we see the largest differences is in order to achieve this, the HD ends up with a 40% hotspot instead of a 65% hotspot that we get for the Millennium.
So that was interesting. We've started to just plan with the arcs with the HD-MLC. We're interested in seeing if we can use HD instead of cones. And then, of course, the thing that always comes up is what's the penumbra. And we have for the HD that the 80% to 50% is the same for the Millennium and the HD. And, of course, a little bit larger 2.4, instead of 1.5, 80 to 20 for the HD and M-120 compared to the cone.
But what we're most interested in is what's the dose fall-off. Instead of the penumbra for an individual field, we're most interested in the dose fall-off and comparing that for the two system cones versus 120. And it turns out, because of the entrance and exit doses, as we know, that it's a bit of a wash and our penumbras for a similar sized target, sphericaltarget, not penumbras, but dose fall-offs are 0.5 to 0.6 millimeters higher for the HD. These are actually results that we got from our RPC Phantom that ended up being lower than what the planning system was giving us. And this is the 80% to 50% distances for in the AP direction left-right, and sup-inf in. And then you can see how for the HD, you're actually lower in these areas and sup-inf, which is typically a longer fall-off because your entering and exit doses is mostly in that direction.
And this was just for fun. We did a plan with a 4-millimeter cone versus a 5-millimeter square. And you end up, because you're using 5 ARCS and 120 degrees, or so for each, roughly a spherical distribution. Not as spherical as the cone, but then the tumors are never really that spherical either. So that's something we're actually going to look into, and next step is to confirm this dosimetrically or verify it dosimetrically. And here are our results from the RPC external Phantom, the head-shaped Phantom with the TLD. The TLD result showed that our delivered dose was 99% correct, and then here they give the offset of the 80% from one edge to the other, the offset of the center of that with the isocenter. And you can see we're pretty 0.4, 0.6 millimeters off in these directions. But we were 1.6 millimeters off in the sagittal coronal.
And this was for a five ARC treatment with the HD-MLC using a fixed frame. And it was a 30Gy treatment, and it took 4.3 minutes to deliver with the 1000 MU per minute setting. So that was encouraging as well, except we're going to look into seeing if there's any systematic error we can remove to improve that.
Conclusions. The system is been nicely stable. We're doing fine with our regular workload. Our distributions are similar or better in the case of, where we have high dose gradients and small concavities, good dose accuracy. These are our commissioning numbers for setting Eclipse up. And hopefully, we'll have more to say about radiosurgery and the HD in the future.
And you probably have seen a lot of these and this as well. Again, the HD-MLC is a lot less rounded than the Millennium MLC and has about 5% more tungsten in it. And we've found that the motor stability on the MLC has been very good. We've had no leaf replacements in six months, and we expect that if we do have one, it'll take a lot longer than a typical Millennium motor since these motors are so much smaller. And we expect them to last longer because the motors are new and improved. The leaf speed doesn't travel at maximum speed at all times, and we don't expect leaf collisions. We didn't have them before at the Millennium, but they have more collision prevention tests during the MLC initialization.
And we've found that with some geometric field shape studies, as well as Picket fence studies and Dynalog studies that are statically position precision is about a 10th of a millimeter. I don't think Varian is quoting this, but that's what we've seen and that the dynamic leaf position precision is one millimeter, according to Dynalog results. Here's a slide on the transmission for the HD-MLC. We have a film on the left, on the right our profiles against and with the leaves. And you can see that the transition for the HD is about 50% lower than the Millennium, and the inner leaf leakage is about half that of the Millennium. And here are our actual transmission values for... And what we've put into Eclipse or configured it with is the 10 by 10 field size value for the transmission, which is about 1 for the 6 MV HD-MLC and about a 1.2 for the 10 MV HD-MLC.
And as you can see, there's a bit of a field size dependence there. Next, we measured the dosimetric leaf gap and got similar results to Fang-Fang. Basically, we use the sliding window technique where we varied the gap of the window and took our chamber measurements for each gap, and then extrapolated to a zero reading to get a dosimetric leaf gap size of 0.8 millimeters. I don't have the units here, but it's millimeters for the 6 MV and about 0.9 for the 10 MV. However, we took it one step further and decided to see if changing our DLG from the sliding window value would improve our dosimetric results so that if we measured the dose distribution for the field, we'd get a better comparison with the Eclipse calculated distribution.
And for that, we used a step pattern, which doesn't have a jagged distribution that might make Eclipse calculation resolution an issue. It doesn't have any tongue and groove issue, and it's not sensitive to positioning. So if we're changing our DLG, we're not compensating for any of these other factors. So basically, what we did, was reconfigure Eclipse with a different DLG value, re-optimized, recalculated the leaf motion. And one point I wanted to make is changing the DLG is not going to change your dose distribution. It's going to change your final leaf motion. Then we remeasured the dose distribution and compared the measured to calculated. And this is the system that we used. Basically, the MapCheck dose calibrated to give absolute dose, and it's at a depth of seven.
And here's our first result for the 6 MV HD. We have measured versus calculated 4.8 sliding window value to a 0.4. And we didn't see much difference. This is an error map on the right for both of them showing whatever is lower of the distance to agreement or the percent dose difference, and the mustard green and the light green here correspond to the 3-millimeter 3%, 3-millimeter distance to 3% dose difference values. So there's not much difference. However, for the 10 MV, when we went from 0.9 to 0.5, we had a marked improvement in our error map.
And we repeated this for a 10 MV for a typical prostate fluence for a single field. And again, you can see the improvement of the 0.5 value for the dosimetric leaf gap used in Eclipse versus the 0.9 we get from the sliding window. And so we did more studies showing the dose difference between measured and calculated for the different steps and for different DLG values and found that we got the best agreement for about the 0.5-millimeter value. And these are, again, our final DLG values. We're also very interested in the HD fluence delivery as well as dose verification. We were somewhat concerned that we might see more tongue and groove with the HD-MLC. So we took several transmission films. Here we have the HD on the top and 120 on the bottom. The tongue and groove here is represented by the white line and happens when one leaf is moving against a static one.
And as you can see, we don't have more white lines with the HD than the M-120. And if we take a profile here, we get an interesting plot where you see that the HD is a lot smoother. This was for a 45-degree angle, and because of the higher resolution is much smoother and has a similar tongue and groove right here, down here. So it does seem to perform much better at the fluence level. We also, for per patient QA, we measured the dose distribution for all the patient fields using the same method I mentioned before, and we've bend our results. So these are for all the fields that we've given our patients for the different systems and energies, the different MLC models and energies. And these are the percent dose difference NDTA we've been reporting to what passed, had 95% of the points pass.
So we see that most of the field dose distribution measurements are here within a 4% dose difference in two DTA, and it's very similar for all of them. So we see that for the HD, we're not doing any worse than we used to, which is really important when your practice is your regular radiation therapy. You want it to at least be the same as it used to be, if not better. So we're not getting anything out here at the higher percent dose difference or higher DTAs. So that was encouraging. Next, We took a look at a lot of dose distributions using the differently sizes. And a lot of people have done that before with different MLC models and different treatment sites. We decided to look at it again anyways and gotten similar results as well as slightly different results. And first, to start off, two sort of similar results. Using the 2.5 leaves versus the 5, we did a regular prostate without nodes, and the distributions are very similar as are the DVHs for the organs at risk. And then similarly, for a complex head and neck with dose painting, 95% around the high dose PTV, 83% around the lower dose PTV. We get similar distribution, a little less hot on smaller leaves and down here.
However, when we looked at our DVHs, we found that the 5-millimeter actually gave slightly better DVHs in this dose range for the brainstem and the cord. But then we got when we went to a much smaller concavities and higher dose gradients, that's when we saw the advantages of the 2.5 millimeter MLC, and our first case was an ethmoid sinus. And here's one of the field fluences. And you can see for the higher resolution HD, how in the small 1-centimeter type concavity, how much better it does than the other models and a block of the other models are, and that translates overall into these distributions. And we have a sharper gradient for the HD 2.5 millimeter. So the 70% doesn't go beyond these lines, whereas the others do. And also, the 10% is curving in here.
And so we're getting more lens sparing, and we see that in the DVHs. And so, in all cases, the brainstem and the lens, we have better sparing with the HD, and for the PTV, we have a more homogeneous dose distribution. And then we did one more. We haven't done any of these cases, but these are my test case, my test PTV. This patient is healthy. There's no tumor. So we did a situation on this side. It's the 5-millimeter. This is the 2.5, and for low cord constraint and low gradient, we get pretty similar results in terms of a hotspot and coverage and gradient. But this changes a lot once we start using a high dose constraint in both situations to end up with a 2.7-millimeter drop-off. And where we see the largest differences is in order to achieve this, the HD ends up with a 40% hotspot instead of a 65% hotspot that we get for the Millennium.
So that was interesting. We've started to just plan with the arcs with the HD-MLC. We're interested in seeing if we can use HD instead of cones. And then, of course, the thing that always comes up is what's the penumbra. And we have for the HD that the 80% to 50% is the same for the Millennium and the HD. And, of course, a little bit larger 2.4, instead of 1.5, 80 to 20 for the HD and M-120 compared to the cone.
But what we're most interested in is what's the dose fall-off. Instead of the penumbra for an individual field, we're most interested in the dose fall-off and comparing that for the two system cones versus 120. And it turns out, because of the entrance and exit doses, as we know, that it's a bit of a wash and our penumbras for a similar sized target, sphericaltarget, not penumbras, but dose fall-offs are 0.5 to 0.6 millimeters higher for the HD. These are actually results that we got from our RPC Phantom that ended up being lower than what the planning system was giving us. And this is the 80% to 50% distances for in the AP direction left-right, and sup-inf in. And then you can see how for the HD, you're actually lower in these areas and sup-inf, which is typically a longer fall-off because your entering and exit doses is mostly in that direction.
And this was just for fun. We did a plan with a 4-millimeter cone versus a 5-millimeter square. And you end up, because you're using 5 ARCS and 120 degrees, or so for each, roughly a spherical distribution. Not as spherical as the cone, but then the tumors are never really that spherical either. So that's something we're actually going to look into, and next step is to confirm this dosimetrically or verify it dosimetrically. And here are our results from the RPC external Phantom, the head-shaped Phantom with the TLD. The TLD result showed that our delivered dose was 99% correct, and then here they give the offset of the 80% from one edge to the other, the offset of the center of that with the isocenter. And you can see we're pretty 0.4, 0.6 millimeters off in these directions. But we were 1.6 millimeters off in the sagittal coronal.
And this was for a five ARC treatment with the HD-MLC using a fixed frame. And it was a 30Gy treatment, and it took 4.3 minutes to deliver with the 1000 MU per minute setting. So that was encouraging as well, except we're going to look into seeing if there's any systematic error we can remove to improve that.
Conclusions. The system is been nicely stable. We're doing fine with our regular workload. Our distributions are similar or better in the case of, where we have high dose gradients and small concavities, good dose accuracy. These are our commissioning numbers for setting Eclipse up. And hopefully, we'll have more to say about radiosurgery and the HD in the future.