Transcript
Thank you. My presentation will partially overlap with the presentation that we had before about the intrafraction movements, but we still found some other values, I hope this will still be interesting.
Now as an introduction, I want to mention again that the CTV to PTV margin should reflect the accuracy of the target localization. And in order to achieve this, we have to optimize both the setup and the internal margins. Now, a long time ago, we improved the setup margin, by the introduction of thermoplastic mask for immobilization for the patient. Now, what we did, we saw between '95 and '99, we treated with the Peacock System in our department. And we already saw that there was a lot of shift and rotation of the patient's head inside the mask system. So we introduced some earplugs with lead beads, and a mouthpiece also with lead beads, in order to improve to patient setup accuracy, so the target localization accuracy. So what we did, we first positioned the patient, based on the references that we had on the thermoplastic mask. And now we actually asked the patient to shift and rotate inside the patient in order to align these lead beads with the lasers. And then this was verified with the double exposure film. And I don't know if you can see it here. In the center of the frame, you can see, actually, the two lead beads almost coinciding with each other. And this was done on a daily basis, very cumbersome I can tell you that. Today we have much more easy tools to perform image-guided radiotherapy. And we are convinced that we also have to do this on a daily basis, in order to reduce both systematic and random errors.
Why we want to do this on a daily basis is also because conformal radiotherapy and IMRT techniques, with the high conformal dose distributions, require a daily setup correction. For instance, if you look at this dose distribution of the skull-based meningioma, you see the high dose level bordering the optic nerve system, and we don't want these high dose levels being delivered to the optic system.
Some advantages of the 6D setup correction is that we have a better agreement between the actual patient position, and the position of the patient during CT scan acquisition. So you have a better agreement of your dose delivery to the patient. Challis already proven us the dosimetric benefit of IGRT with online 6D setup correction, and the PTV margins that we use these days account for setup accuracy, which is much improved with the 6D setup, the intrafraction motion, and the target delineation uncertainty. Nowadays, we use a uniform margin of 3 mm for intracranial treatments in our department. Now, when we look at frame-based positioning, we see that the patient is immobilized with the relocatable mask, two layers, a front part and a back part. Localization is performed with this localizer box using these three target positioner overlays. So, only based on the three lasers and we do not have any information about rotations and the shifts of the patient inside the mask system.
Now Robar evaluated this, he performed 44 offline CT evaluations of 8 patients, with the relocatable mask system with a U-frame, and he detected shifts ranging from minus 2.2 to plus 3.5 with a large standard deviation. The shifts in the lateral and vertical direction were much smaller, but he also detected rather large rotation. So you can see the longitude rotations around the longitudinal axis being wrong, range from minus 2.8 to 2.6. And also rather large rotations around the lateral axis and the vertical axis were detected. And when we used the thermoplastic mask with the localizer box, these are not corrected prior to treatment.
So we have some interest to go frameless. So, in our department patients are also immobilized with a similar thermoplastic mask. We use a personal infrared marker configuration which we place on the mask, so we do not use the array. And the reasons why we use a personal configuration is because of safety reasons that we initiate that we introduced in our department because the patient list of the exit track system is not linked to the patient list on the verification system of the linac. So when we use a personal configuration, we are sure that the correct patient is inside the mask for the correct treatment.
When the patient is positioned we perform target localization based on stereoscopic X-ray imaging, 6D fusion all the rest. So we have 6D information of our patient position inside the mask system. So we started patient treatments frameless in 2007, and I evaluated 49 patients, all treated between February last year and December last year, the age of the group ranged from 8 years to 79 years with an average of 57 years, and equal distribution between male and female. And the patients were either treated with non-coplanar dynamic conformal arcs or with non-coplanar IMRT treatments.
So the patients got a 6D setup correction, a verification of this setup correction, and then a verification at the end of the treatment. And it was actually the difference between these two verifications, that gave us information of the intrafraction motion of the patient inside the mask. A total of 1302 fractions, the fraction per treatment course ranged from 5 to 35. And 97% of the verifications were retained for analysis. The time interval that we had for the intrafraction motion was an average 14.6 minutes range from 5 to 34 minutes, an average standard deviation of 2 minutes and a group standard deviation of almost 4 minutes.
Now, if you look at the results of the setup shift distribution, so we see here, the cloud of all these measurements that are projected to the planes. If you look at the mean values, we see almost zero for longitudinal and lateral, but we see a 1.3 mean value for the vertical direction. And this is actually caused by the fact that we put the infrared markers to the mask system, and not to the array. Because when for instance, the patients are taking corticosteroids during treatments, the faces are swelling, and then we put thicker spaces between the two layers of the mask. And this induces, of course, a vertical deviation of the positioning of the patient. But for the three directions, we see rather large standard deviation. All the displacements is less than 2.76 mm in 95% of all fractions. And if we use the famous Van Herk formula, we see that the margins that we would need in our treatment planning were about 4 mm in all directions.
So we do have to correct these setup shifts based on the patient information inside the mask system.
A short look at the 3D vector. So we see that the 3D vector of the setup shifts is mostly in between 0 and 3 mm, a mean value of 2.4, standard deviation of 1.2 . The maximum that we've found was up to 9 mm. The 3D vector was...for the lateral and longitudinal, there was no preference in any direction. In the vertical direction, there was, of course, a preference in the anterior direction due to the fact that we put the infrared markers on the mask like I explained before. If we look at a set of rotations, we see almost...several for the mean values of the longitudinal and the vertical rotation, but we see a large mean value for lateral rotation. And this is due to the fact that the patients cannot hold their heads like the mask is molded. So the lateral rotation is like the pitch. And it's very hard to keep that patients head in that same position because the patients always tend to flex their heads down. This is resulting in this large value over there. You also see some large standard deviation, and the rotations are less than 2.68 degrees in 95% of the fractions.
An overview here of the frequency distribution, so we see that only...a little more than 50% of the rotations are smaller than 1 degree. And we see a rather large range in the detected rotations in the setup. The orientation of the rotations, like I explained, we see a preference in the negative lateral direction for the rotations. The other two don't show any clear preference in rotation direction.
This brings me to the intrafraction shift distribution. If you look at the mean values, we have almost zero for all directions, much smaller standard deviation, and also much smaller ranges of the measured intrafraction motion. The displacement is less than 1.48 mm in 95% of all fractions, and if we now use the famous Van Herk formula again, you see that we get values that are within the 3mm millimeter that we use in our treatment planning.
The 3D vector for the intrafraction is mainly less than 2mm, mean 3D vector of almost 1mm standard deviation 0.5, and the maximum 3D vector that we detected intrafraction was 3.6 mm. The orientation of the 3D vector in the intrafraction shifts has never a preference in any direction. The intrafraction rotations, again very small values for the mean values, also, a large reduction in the standard deviation, the range is between -2.0 and +2.0 degrees. So the rotations are less than 0.8 degrees in 95% of all fractions. So once the patient, the setup is corrected, the patient is fixed inside the mask. Here in the frequency distribution, you see that all rotations, or almost all rotations, we have a small amount of gauze over it, but almost all rotations are less than 1.0 degree, mean value approaching zero, standard deviation of point 0.4, and also for the rotations, we see that there is no preference in any direction.
So in conclusion, the results of the setup measurements indicate the necessity to correct for the patient shift and the rotation inside the mask system in order to increase the treatment delivery accuracy. And the application of the 6D setup correction gives us a better setup accuracy than the frame-based approach.
And then the intrafraction shifts and rotations are all very small. More than 95% of the measurements are within the tolerance levels that we use of 2 mm and 1.0 degree. And there was a high immobilization capability of the frameless mask system, even for the longer treatments. And therefore I want to thank all my colleagues who have helped me with this study. And that's it.
Now as an introduction, I want to mention again that the CTV to PTV margin should reflect the accuracy of the target localization. And in order to achieve this, we have to optimize both the setup and the internal margins. Now, a long time ago, we improved the setup margin, by the introduction of thermoplastic mask for immobilization for the patient. Now, what we did, we saw between '95 and '99, we treated with the Peacock System in our department. And we already saw that there was a lot of shift and rotation of the patient's head inside the mask system. So we introduced some earplugs with lead beads, and a mouthpiece also with lead beads, in order to improve to patient setup accuracy, so the target localization accuracy. So what we did, we first positioned the patient, based on the references that we had on the thermoplastic mask. And now we actually asked the patient to shift and rotate inside the patient in order to align these lead beads with the lasers. And then this was verified with the double exposure film. And I don't know if you can see it here. In the center of the frame, you can see, actually, the two lead beads almost coinciding with each other. And this was done on a daily basis, very cumbersome I can tell you that. Today we have much more easy tools to perform image-guided radiotherapy. And we are convinced that we also have to do this on a daily basis, in order to reduce both systematic and random errors.
Why we want to do this on a daily basis is also because conformal radiotherapy and IMRT techniques, with the high conformal dose distributions, require a daily setup correction. For instance, if you look at this dose distribution of the skull-based meningioma, you see the high dose level bordering the optic nerve system, and we don't want these high dose levels being delivered to the optic system.
Some advantages of the 6D setup correction is that we have a better agreement between the actual patient position, and the position of the patient during CT scan acquisition. So you have a better agreement of your dose delivery to the patient. Challis already proven us the dosimetric benefit of IGRT with online 6D setup correction, and the PTV margins that we use these days account for setup accuracy, which is much improved with the 6D setup, the intrafraction motion, and the target delineation uncertainty. Nowadays, we use a uniform margin of 3 mm for intracranial treatments in our department. Now, when we look at frame-based positioning, we see that the patient is immobilized with the relocatable mask, two layers, a front part and a back part. Localization is performed with this localizer box using these three target positioner overlays. So, only based on the three lasers and we do not have any information about rotations and the shifts of the patient inside the mask system.
Now Robar evaluated this, he performed 44 offline CT evaluations of 8 patients, with the relocatable mask system with a U-frame, and he detected shifts ranging from minus 2.2 to plus 3.5 with a large standard deviation. The shifts in the lateral and vertical direction were much smaller, but he also detected rather large rotation. So you can see the longitude rotations around the longitudinal axis being wrong, range from minus 2.8 to 2.6. And also rather large rotations around the lateral axis and the vertical axis were detected. And when we used the thermoplastic mask with the localizer box, these are not corrected prior to treatment.
So we have some interest to go frameless. So, in our department patients are also immobilized with a similar thermoplastic mask. We use a personal infrared marker configuration which we place on the mask, so we do not use the array. And the reasons why we use a personal configuration is because of safety reasons that we initiate that we introduced in our department because the patient list of the exit track system is not linked to the patient list on the verification system of the linac. So when we use a personal configuration, we are sure that the correct patient is inside the mask for the correct treatment.
When the patient is positioned we perform target localization based on stereoscopic X-ray imaging, 6D fusion all the rest. So we have 6D information of our patient position inside the mask system. So we started patient treatments frameless in 2007, and I evaluated 49 patients, all treated between February last year and December last year, the age of the group ranged from 8 years to 79 years with an average of 57 years, and equal distribution between male and female. And the patients were either treated with non-coplanar dynamic conformal arcs or with non-coplanar IMRT treatments.
So the patients got a 6D setup correction, a verification of this setup correction, and then a verification at the end of the treatment. And it was actually the difference between these two verifications, that gave us information of the intrafraction motion of the patient inside the mask. A total of 1302 fractions, the fraction per treatment course ranged from 5 to 35. And 97% of the verifications were retained for analysis. The time interval that we had for the intrafraction motion was an average 14.6 minutes range from 5 to 34 minutes, an average standard deviation of 2 minutes and a group standard deviation of almost 4 minutes.
Now, if you look at the results of the setup shift distribution, so we see here, the cloud of all these measurements that are projected to the planes. If you look at the mean values, we see almost zero for longitudinal and lateral, but we see a 1.3 mean value for the vertical direction. And this is actually caused by the fact that we put the infrared markers to the mask system, and not to the array. Because when for instance, the patients are taking corticosteroids during treatments, the faces are swelling, and then we put thicker spaces between the two layers of the mask. And this induces, of course, a vertical deviation of the positioning of the patient. But for the three directions, we see rather large standard deviation. All the displacements is less than 2.76 mm in 95% of all fractions. And if we use the famous Van Herk formula, we see that the margins that we would need in our treatment planning were about 4 mm in all directions.
So we do have to correct these setup shifts based on the patient information inside the mask system.
A short look at the 3D vector. So we see that the 3D vector of the setup shifts is mostly in between 0 and 3 mm, a mean value of 2.4, standard deviation of 1.2 . The maximum that we've found was up to 9 mm. The 3D vector was...for the lateral and longitudinal, there was no preference in any direction. In the vertical direction, there was, of course, a preference in the anterior direction due to the fact that we put the infrared markers on the mask like I explained before. If we look at a set of rotations, we see almost...several for the mean values of the longitudinal and the vertical rotation, but we see a large mean value for lateral rotation. And this is due to the fact that the patients cannot hold their heads like the mask is molded. So the lateral rotation is like the pitch. And it's very hard to keep that patients head in that same position because the patients always tend to flex their heads down. This is resulting in this large value over there. You also see some large standard deviation, and the rotations are less than 2.68 degrees in 95% of the fractions.
An overview here of the frequency distribution, so we see that only...a little more than 50% of the rotations are smaller than 1 degree. And we see a rather large range in the detected rotations in the setup. The orientation of the rotations, like I explained, we see a preference in the negative lateral direction for the rotations. The other two don't show any clear preference in rotation direction.
This brings me to the intrafraction shift distribution. If you look at the mean values, we have almost zero for all directions, much smaller standard deviation, and also much smaller ranges of the measured intrafraction motion. The displacement is less than 1.48 mm in 95% of all fractions, and if we now use the famous Van Herk formula again, you see that we get values that are within the 3mm millimeter that we use in our treatment planning.
The 3D vector for the intrafraction is mainly less than 2mm, mean 3D vector of almost 1mm standard deviation 0.5, and the maximum 3D vector that we detected intrafraction was 3.6 mm. The orientation of the 3D vector in the intrafraction shifts has never a preference in any direction. The intrafraction rotations, again very small values for the mean values, also, a large reduction in the standard deviation, the range is between -2.0 and +2.0 degrees. So the rotations are less than 0.8 degrees in 95% of all fractions. So once the patient, the setup is corrected, the patient is fixed inside the mask. Here in the frequency distribution, you see that all rotations, or almost all rotations, we have a small amount of gauze over it, but almost all rotations are less than 1.0 degree, mean value approaching zero, standard deviation of point 0.4, and also for the rotations, we see that there is no preference in any direction.
So in conclusion, the results of the setup measurements indicate the necessity to correct for the patient shift and the rotation inside the mask system in order to increase the treatment delivery accuracy. And the application of the 6D setup correction gives us a better setup accuracy than the frame-based approach.
And then the intrafraction shifts and rotations are all very small. More than 95% of the measurements are within the tolerance levels that we use of 2 mm and 1.0 degree. And there was a high immobilization capability of the frameless mask system, even for the longer treatments. And therefore I want to thank all my colleagues who have helped me with this study. And that's it.