Setup Accuracy and Efficiency of Real-Time Video Assistance for Surface-Guided Positioning in Radiation Therapy of Right-Sided Locoregional Breast Cancer

Professor Taran Paulsen Hellebust (00:04):

So first, a few words about the implementation of surface guided radiotherapy at Oslo. But this resulted in us being a University Hospital. Our hospital has 16 linear accelerators at two sites. And in 2017, the first align artist system was installed. And gradually, the system was then installed in all linear accelerators. In 2020 and 2021, we actually conducted a study where we compared the setup accuracy and also efficiency with the system to the skin marks tattoo. And this is not the topic of this lecture, but still I think it’s important to include it here, as it is relevant. And that study, which was published in 2022, showed that the system had higher accuracy compared to tattoos and skin markers, but we did not find any time gain for this.

Professor Taran Paulsen Hellebust (01:17):

And then in September 2021, the AlignRT Advance system was installed at one of our sites. And included in that upgrade, we had a test license for the Postural Video function. And prior to that upgrade, when we aligned the patients, we used these bars that I’m sure you all are familiar with. They are showing the position, the real position, the real time position of the patient compared to the surface from the planning CT inside this ROI here, as you can see and in this example, we can see that the arm position is not well, and to be able to control that, we had to take this treatment capture compare a line, if it’s not here, it’s not proper.

Professor Taran Paulsen Hellebust (02:18):

So we had to, then again, ask the patient to move their arm and then take another treatment capture and then go on to, to repeat that until it fits well. And then afterwards, we are using this as a Postural Video function that I’m also sure that many of you are familiar with, where there is a real-time video included, and you can constantly compare that to the surface from the planning city. However, the disadvantage of such a test license, I have to say, is that it’s expiring. And it’s also something that you have to purchase. So, it was quite expensive since we have so many Linacs. So we decided that we had to do an evaluation of this functionality. And then we developed a rather small survey that we distributed to 26 RTTs that has been working on the Linac with the, the functionality.

Professor Taran Paulsen Hellebust (03:20):

And I think from the replies on the survey, it was quite clear that there was a strong opinion that the Postural Video function reduced both the treatment time and also increased the accuracy. And this is just an example from one of the questions where we asked, how important is the postural video during patient alignment? And as you can see here, the rep, the responders, they had to grade on a scale six point scale from not very important to very important. And I can see a clear least priority here, answer that. It was very important. However, I have to say that this small survey was not sufficient to convince our management team to invest in this. So they wanted more solid data. So that’s why we made a business case where we found out that if we could show that using this functionality could reduce the set of time by around one minute per patient, we could actually reduce our activity or the, the opening hours to say like that with one evening shift.

Professor Taran Paulsen Hellebust (04:39):

And as you know, an evening shift is usually more expensive compared to a day shift. So this was the, the, the gain we wanted to see, sorry. So that’s why we then decided to run a prostitute trial comparing the setup procedure with and without this postural function. So we chose to use the same design as we had in the previous study that I just referred to. That’s called a randomized crossover design. And that means that all the patients will be aligned with both of the procedures. It means that one has the postal video function and the other has this treatment capture procedure, and since all the patients are aligned with both the procedures, it means that the patient will be their own control, and thereby statistically we can, we don’t have to include that many patients.

Professor Taran Paulsen Hellebust (05:40):

We also decided to minimise the effect of the fraction number. We decided then to randomize which order of the procedure to do for each patient. And we also decided not include the data on the first fraction since we are also doing some timing. And here we inform the patient, and some other things are done. So that was excluded from the analysis. So from a poor ana poor calculation, we found out that we needed to include 26 patients. We wanted to have a quite clean cohort of patients. That’s why we didn’t want to include patients with deep inspiration breath holes. And at that time, we did that only for left-sided patients. That’s why we in this study, only included right sided patients. We also wanted to control and to evaluate the position of the lymph nodes. So we only included patient also with regional nodes included in the target volume.

Professor Taran Paulsen Hellebust (06:49):

So at the treatment, we aligned with these two procedures or set up with two procedures, and then we performed a cone beam CT that was done every fraction. And we then corrected the position of the patient with a six-degree table. So it means that the treatment was actually based on the CBCT and not the pure surface guided alone. However, we use this shift from this match to calculate the potential systematic and random errors if the alignment had only been done with the surface guidance. And then we did a time recording. We had a dedicated person who was taking the time not involved in the treatment as such. And then the RTTs during the procedure, then said now, and then the person was writing down or using a stopwatch, actually.

Professor Taran Paulsen Hellebust (07:55):

And you can see here the different time points we recorded from the first, when the patient walks through the doors of the treatment room, patient, horizontal on the couch with the arms up, serve on patient treatment, in treatment position, and then start of CBC acquisition and then the radiograph press button beam on. And then the last time point was when the patient exited the treatment room, and then we could calculate all kind of different timing. And this is the one which we felt was most important. And as I said, for the fractions that the patient was aligned with, without postural renal function, we used this predefined region of interest here, as you can see indicated here. That was including the stable chest wall, and we excluded the arm, the shoulder, and the breast tissue, so to be sure that this was quite, quite stable.

Professor Taran Paulsen Hellebust (08:59):

And then to evaluate the position of the lymph node. So the accuracy of that part. We imported all the CBCs into our treatment planning system, which is raised station. And then we’ve performed two co-registrations, and we used the clavicle as a surrogate for the lymph nodes. So the first co-registration was between the CBCT and the planning CT. And then we used the shifts from the online match. So that’s actually giving us the, the real position of the clavicle during treatment. And the next one we used again, but we, we, we coated the CBCT to do the planning CT. And now, but now we used the clavicle as a matching ROI, so that means it was a near perfect match between the clavicle on the sea, B, C, and on all the sea B cities on the planning city. And this meant that we can then produce such images where I think it’s very, it’s very difficult to see, but beneath her, there’s a gray structure, which is showing the position on the planning C or the clavicle. And then you can see the green one is the position during a non Postural Video function alignment at the red one when we use it. And then we use the dial similarity cent to, to quantify the accuracy of these two different set procedures.

Professor Taran Paulsen Hellebust (10:33):

So here come the results. This is a box plot showing the, sorry, this is the patient along the X axis here, and I can see it is blue for the fractions without the postural video function and red with dysfunctionality. And this is just a plot showing the shift in the vertical direction. But we did that, of course, in all other directions and also with rotation. And as you can see here there is actually a significant difference. We used a A-P-R-D-T test to control that. We also used some other statistical tests, but that’s what we, we used for this p-value. And we can see that the variation is quite considerable between the patients, and also the difference seen here, even though it was statistically significant, it was not clinically significant or relevant.

Professor Taran Paulsen Hellebust (11:34):

So we did not see any other P values below 0.05 for any of the other directions or rotations. And then this is in the, the dice similarity and also we I forgot to say, but we’ve also calculated the Hof star distance. 95% of that we did not fund any differences between the two procedures. And here you can see quite clearly that there’s a big variation between the different patients here with regard to, for instance, this one has a quite low and a quite large spread for the diocese for both procedures actually compared to, for instance, this one.

Professor Taran Paulsen Hellebust (12:18):

However, when it came to the timing, we could find some differences. This is again the same type of plot. We can, we actually found that the time number T two, if you remember it, was from the patient who was horizontal on the couch with the arms up until the patient was aligned. We could find an average reduction of 40 seconds, which was clearly statistically and also clinically significant. So this means that the conclusion is that by using the postural video function during the setup of right-sided local regional breast cancer patients, we can significantly reduce the set of time with com without compromising the set of accuracy. So the business case was approved <laugh>, and we were allowed to invest in the possible video function on all oral Linux.

Professor Taran Paulsen Hellebust (13:23):

Additionally, the opening hours were reduced with one evening shift on one of the Linac and the work was published. So if you are more interested, you can go into this publication that was published during the springtime. I have to say that this, it was a multi-professional team that did this work. None of these physicists was able to come here today. I have to admit that I haven’t been working clinically for some years, but I was the, the I designed the study and also we had RTTs involved here. So it’s really a very nice team working on this kind of project. And now we are looking into another type of project that we will do in the future. So thank you very much for your intention, and I’m willing to answer questions if I’m able to answer them.