Michael Tallhamer
Chief of Radiation Physics
AdventHealth Parker, USA

Video Transcript

Michael Tallhamer (00:03):

Good morning everyone. I guess it’s morning there. Evening here. I’m Michael Tallhamer. I’m the chief physicist for an organization called AdventHealth in the Rocky Mountain region. I’m based here in Denver, Colorado which is a beautiful place, but I’m unfortunately not able to be with you today. I’d love coming to your area of the world. One of my favourite places, and I’m just disappointed I couldn’t get there this year. But I’ll be talking today at least in this first talk, about changing our perspective on SGRT expanding what we call the SGRT Umwelt with DoseRT, which is Cherenkov Imaging Based Technology. We’re going to discuss some of our early findings, some value propositions for the technology and some clinical use cases. Cherenkov imaging is I think, is fairly well understood what you see, if you could see the background of my slide here I don’t know if sometimes they subtract them out, but if you’re seeing two particles passing each other what you’re seeing is a finite element simulation of what Cherenkov imaging is. It’s essentially a particle travelling faster than the speed of light in a given medium, at some point in the middle of the screen there, you’ll see a shock wave produce, and the emission spectrum is the yellow orange that you’re seeing there. And the DoseRT product is taking advantage of this effect, and it’s a technology that we’ve had within our clinics here now for a little over a year and a half here in Colorado, and then at one of our Florida facilities in Celebration Florida which is just outside of Orlando AdventHealth has a second system now of DoseRT, and we’ve been working together with the celebration team to put together some case studies and also with some of the people within Cherenkov what they’re calling Cherenkov Consortium or Imaging Consortium looking at different applications of Cherenkov imaging. Before we start, these are my disclosures. Parker Colorado has PSA with Vision RT. We have a COE site, which is our celebration facility, which happens to have the DoseRT as well. And then I am a physics consultant. I provide physics consultation services to a variety of different vendors. These two vendors are typically the two vendors that you will see in any of my talks for SGRT. So those are listed there. And then credit for this simulation goes to Nils Bergland, who produced the finite element analysis to show how Cherenkov imaging takes place. I tend to and tell someone a particle travels faster than the speed of light in a room full of physicists, sometimes I get the side eye. So I’m, this is just proof if the math works out.

Michael Tallhamer (02:38):

AdventHealth we’re a fairly large organization. We’re 52 campuses across nine states. 11,000 plus beds at this point. This is a little bit older slide, and as I said, two facilities, one in Colorado, which is where I’m at today. And then one in Florida that just has this, I believe they’re going on about nine months of having the system as well and doing incredibly good work with the system down there in Florida, the technology comes out of Dartmouth, which is in New Hampshire, which is up here in New England. And some of the cases you’ll see here are publications from Dartmouth as well, but the majority will be just some of our clinical cases and what we’re doing with the technology within AdventHealth itself.

Michael Tallhamer (03:26):

Before we get started to define the term Umwelt, the reason why I like this term is it’s an organism’s unique sensory perception of the world. And it depends on how it detects and interprets that. It used to be something that’s common in medicine that we’re familiar with the idea that we’re made of these large, biologically significant molecules. It’s now part of kind of everyday common practice society, at least Western societies are very well versed in this. We’re bathed in this. We have commercials about this stuff all the time. But what we often as scientists look at is those large biological molecules are made up of smaller constituent molecules or smaller constituent atoms. This is the very first published photo of a hydrogen atoms orbital structure of electrons using what’s called quantum microscopy. We have expanded our Umwelt as scientists and clinicians to look down into ever and ever more finer structures of the objects we’re looking to look at. And so we can see things down to these orbital structures. I think this is actually the small structure. I think we’ve been able to image, truly image we can detect subatomic particles, but to actually image something. I believe this is one of the smallest things. And then if we are used to looking into telescopes, I love going out to your neck of the woods, New Zealand and Australia, and getting kind of away from the cities and being able to see the stars in the Milky Way. It’s something that I’ve always liked doing as a scientist and someone who used to take you know, pride in being in deep space imaging physics. We are acutely aware that we’re embedded in a system of much larger things, but our perception is unique in that we have a very narrow band of perception. And our narrow band of perception allows us to see the world in, you know, just one level of detail. And we create tools that allow us to see it in other details. And that’s what we’re doing with our and SGRT.

Michael Tallhamer (05:18):

SGRT Umwelt is being expanded by using a technology similar to what I described to people as the Hubble Deep Field experiments that we did in the early two thousands. We took the Hubble telescope, we pointed it at what we called an empty portion of the sky that area of the sky as far as size-wise, if you held up a pencil at arm’s length and pointed it to the sky, the size of the area of the sky we were looking at was about the size of the pencil lead. And we exposed the Hubble telescope over and over and over and over and again to this area of the sky. And then we, those images after a number of months, what we found out, that empty area of the sky was not as empty as we expected. And everyone at this point, because I can’t see the crowd is probably wondering, like, I think this guy’s at the wrong conference. This has nothing to do with SGRT. But this is exactly what we’re doing with DoseRT. We are essentially pointing cameras with image intensifiers at an area isocenter essentially, and the areas around the isocenter and collecting the photons that are emitted from our patients for free during the radiation delivery process. And now we’re trying to process that image and use some insights from that data collected.

Michael Tallhamer (06:18):

So the goal of this presentation is more or less to share the example use cases of Cherenkov imaging focusing more on clinical applications of the software as it’s currently encompasses the quality and safety applications, the visualization of stray dose that we’re finding. And then some of the plan robustness evaluations that we’re doing, and the implications of how Cherenkov imaging can be used to recommend clinical changes in our practice. We’re also going to discuss the difference between the data that we’re collecting and the insights that we gain and the challenge interpreting this kind of data from a clinical perspective on a day-to-day basis, trying to make heads or tails out of what we’re looking at.

Michael Tallhamer (06:56):

So we’re going to delve right into quality and safety, and you’ll see a lot of these cases are going to cross over into a number of these categories. I’ve broken them up just simply because it makes it a little bit easier for us to kind of look at different aspects of what we’re looking at. The basic value proposition for us is that you can see the delivery of the dose. And so what we see here in the upper left is a standard 3D T spine. We have an upper esophageal case in 3D. This is an IMRT VMAT partial breast. This is a VMAT delivery of a recurrent head and neck. We’re looking at an IMRT static field, IMRT, delivery of a nodal SBRT, and then a VMAT delivery of a flank sarcoma. So you can see the visualization spans a large variety of different body sites at different techniques and things that we’re doing. And so the utilization as far as looking at the geographic placement of the dose has some utility just in of itself. That’s our basic value proposition.

Michael Tallhamer (07:54):

If we look at this, this is a case that I, this is one of the first cases I’ve ever presented. I believe I even presented this last year when I was talking about early findings. This is the only technology where you can take a plan and look at the region where the dose should be. This is the live view that the dosimetrist is seeing during delivery. So you can see the deposition of the dose live in video format, and then come back to this dose and look at this and evaluate that dose for a ppropriateness to see if there’s any challenges. That’s the dose going where we would expect it to be and start using this technology as a way to solve problems. So in this case, this is a 36-year-old female. If you were really astute and looked at the very first few frames of the video, there was this flash of dose across both breasts. This was actually the most common error found after we implemented the technology, was the fact that this is a manually selected port film technique. They selected the pelvis port film technique by mistake. It’s one entry above the breast port film technique, and it was just an errant click inside of Aria. We’ve been able to identify this. This is one monitor unit delivered through a 22 through 40 field. And we were able to find this and then put administrative controls in place to, to eliminate this. There’s no hard stop for this. We thought we had eliminated this air from our centers. And then many months ago I think three or four months ago at this point, maybe five months ago at this point, we upgraded to TrueBeam 4.1. And when we upgraded to TrueBeam 4.1, this problem mysteriously started appearing on all of our patients again. And we thought, well, maybe this is out of practice. We have new therapists, maybe something went down. We started looking into it and found that there’s actually a bug in the 4.1 upgrade for the TrueBeam, and that it actually ignored the jaw settings for all of our port film techniques. So all of our port films were being opened to 22 by 40 in the case of a HDMLC linac. And so we were able to report this to Variant, but this was immediately caught on the very first day after the import. And we had done all of our testing. We had done all of our validation and commissioning over the weekend, but we had not taken ports on anything because we typically don’t port film our phantoms. And so this was immediately caught, identified, and reported just by using the Cherenkov imaging system.

Michael Tallhamer (10:00):

This is a 67-year-old female prone breast. Prone breasts are pretty common error centers. The first two fractions look like what you see here on the left. Nothing to really note decent dose distribution, exactly what we’d expect to see from the treatment plan. And then on fraction three, we found this. And this ended up being a patient who has back pain issues. She doesn’t have a bad back, but she back pain issues and in an effort to help support her weight on her shoulders rather than on her back, because she’s relatively stretched out, if you look in this view, she pulled her elbows down and actually dislodged this cushion from the index bar. You can see some dose Cherenkov imaging dose that’s coming off the pad and then also exiting dose through her upper arm. This was identified immediately by the therapist after treating the medial field. We were brought in as the physics team went into investigate what was going on, found the issue, helped her out, supported her back, got her back to being comfortable, and was able to successfully deliver the rest of the treatment without any issues. This was corrected for all subsequent fractions. And so what we’re finding is that most of the issues that we’re finding with this technology are in the first one to five fractions. The errors that they catch tend to be planned robustness issues or body habitus or compliance issues. And we can address those very early on before we start seeing problems.

Michael Tallhamer (11:15):

A good example of this, this is day one of a high tangent breast. This patient was verbally given instructions multiple times to raise her chin because she’s getting these high tangents postural video immediately identified that her head was in the wrong position. She refused to comply. And then after the first field, we were called to the machine, looked at the composite image and sure enough, we can see dose on the patient’s chin. What we weren’t expecting to find was this error. So because she’s doing this body crunch and because of her body habitus and this role of bubble wrap, she’s actually, as she’s doing this body crunch, she’s pushing breast tissue outside the inferior border of the field. And so while we were looking to address this, we actually found another plan robustness problem. And so this was not apparent in the plan. This error, obviously, she was stretched out in the plan. She wasn’t crunched up like this. And so we actually adjusted the inferior field border to include all of the breast tissue and some margin just simply because we could not ensure that she would be compliant for all subsequent fractions. So we increased the robustness of the plan. We addressed this with the patient. The physician actually wanted to show her the consequence of tucking her chin. She was able to see what it was doing, and she was very compliant after that as far as raising her chin and keeping it out of these high tangents. And so multiple factors identified. Some we anticipated others we didn’t, but, and some we were able to address from a planning perspective.

Michael Tallhamer (12:38):

Stray dose visualization is something that we see often in our clinics. Not often that we see it, but we’re also looking for it. So we’re trying to create plans that will not have that. And you’ll see kind of consequences of that here in a moment. This is it out of the original publications that came out of Dr. Jarvis’s group showing that the AlignRT tolerances were within tolerance, but this patient who could not raise their arm showed even though they were intolerance, some contralateral breast dose on two days, some ipsilateral shoulder dose on a couple other days, showing that maybe your SGRT tolerances, if they’re statistically derived, are only good for a certain set of population of patients, and that you may have patients that live in the tails of those distribution. A good example of that, just some, some general errors, not necessarily a AlignRT tolerance issue is, in this case from our Florida facility. These two on the right one is it’s just skimming the contralateral breast. And you can see some dose to that contralateral breast with small degrees of roll. Here’s another planning problem where the angulation of the treatment field treating this photon cavity boost after whole breast radiation was not angled in a way. And it’s actually skimming the contralateral breast. And this had to be addressed from a planning perspective. And then this in the middle is one that I have shown in the past. This is one of our very first patients who had a PAB in super clave delivering this field super clave. This was her first day. She had some pretty brisk reactions under her arm. And we were trying to figure out in her last few fractions what was causing this because. All of her port films look great. Everything looked fantastic. Her SDRT tolerances were always in. And on the second day, we monitored her. We saw this, and this is a function of a small half a degree of roll and the steep slope of the reconstruction of her chest wall with this expander. So a small degree of roll, even though we only allow for one degree that half a degree was old enough to allow for this to flash onto that steep slope. And the other interesting factor is this is a midpoint control point. So if we were to look at this using the light field before delivery, if she was rolled, the first control point was smaller than this, so we wouldn’t have seen this even with the light field. And so this is another indication where Cherenkov imaging is actually providing a glimpse into these mid treatment control points, which we’ll see in a little bit of how the significant those midpoint control points could be. But those are not accessible to us at the start of delivery.

Michael Tallhamer (14:59):

This is another case like that. This is from our Florida facility extremely large treatment volume, standard chest wall, we thought, and then comes back with this extensive volume and involving skin, going all the way up into the underarm and treating contralateral node volumes. It was decided that they wanted to treat this VMAT because of the large shape and kind of disparate regions of delivery that needed to happen. If you look at the CT is really nothing of node CT started and stopped, kind of where we always would expect. But if you’re looking closely, you can kind of see what may be happening here.

Michael Tallhamer (15:32):

On the very first day, on an effort of that VMAT to treat to this high arm volume, you can see it’s going through the arm, but then exiting out the shoulder, entering the face, and exiting out the contralateral side of the face, the CT ends at the chin. So this was not something they saw at the treatment plan. But immediately on the first day of treatment, it was evident that there was dose that was exiting the face. Another midpoint control point shows a significant dose up underneath the CHA or cheekbone. And so again, something that was not accessible in the treatment plan, but very obvious on DoseRT, when you’re looking at the Cherenkov imaging.

Michael Tallhamer (16:10):

This is one where we’re prospectively looking to avoid the ipsilateral limb. She has a frozen shoulder and frozen elbow. So her arm is permanently in this position. We had to optimize to a partial bolus field, so she was simed with this bolus. So we’re concerned about making sure the bolus is in the right position because the optimized fluence accounts for that bolus position. And so we can watch during delivery to make sure we don’t see dose on the ipsilateral arm. We can look at the composite imaging to make sure that there’s no additional dose on the arm. We can look at a specialized rendering of the CT that we can do and look at basically, this is a projection of the surface dose onto that rendering to see, you know, roughly where we can expect to see Cherenkov signal.

Michael Tallhamer (16:52):

And then we can also blend this using some camera magic and some Python code. We can look at where her ear lobes, her chin, her ipsilateral and contralateral shoulders, her ipsilateral arm, all of these things are being lined with postural video. So we would expect a very high degree of coincidence between all of these structures. And so we can continue to monitor this proactively every day to make sure that we’re delivering this static field IMRT in an efficient but effective way as well.

Michael Tallhamer (17:19):

And that brings us to plan robustness. Plan robustness is something that we are really stressing because of this technology we’re seeing little things that can be tweaked and made better because of the Cherenkov imaging on our patients. And so we’re actively trying to look at how to translate the knowledge of the complexity of what we’re doing in a plan to the therapy staff who’s actually delivering this. Because we’re not just taking into account structures that are at isocenter, we’re taking into account the entire patient’s anatomy in many cases, using things like avoidance sectors and things like no entrance, no exit structures. And those things need to be in the same position as the time of sim. So they get the benefit of those optimizations at the time of delivery.

Michael Tallhamer (18:00):

And so we look at something here, this is an earlier version of the software. So you’ll see a single view, and there’s two camera views here. And so you’ll see the therapist kind of clicking between the two. And this is postural video down here. So very small, hard to see. But this is no longer the version that we have. But they’re oscillating back and forth to make sure they can see dose, but also that they’re not seeing contralateral limb dose. And we were doing this to kind of teach them, because this plan is a thigh sarcoma treated with VMAT. And we could not get this limb out of the path of the arcs. Obviously, physician didn’t want to do static field. The patient had a bunch of other contraindications, very high pain, a lot of other things, wanted this delivery to be as quick as possible, wanted VMAT. And so we were told them like, we’re, we’re going to be using avoidance sector, please, you know, make sure that the contralateral limb is accounted for. After 30 fractions as a kind of a post audit, we found four fractions where there was some or some significant in this case, contralateral limb dose. And the reason for that is as we planned it with the, the legs kind of scissored out in this kind of frog legged setup we are using an avoidance sector. And the avoidance sector is pie-shaped. So it’s taking a sector out of the arc. And as these legs come back together, there’s those legs come back together, you’re moving that thigh into the narrow area of that avoidance sector. And if that’s not communicated to the therapist and they’re focused solely on the ipsilateral affected limb, you could easily move a, a structure that was used during optimization to avoid, but move it into an area where it’s receiving dose. And we found four fractions where we saw at least some detectable dose on the contralateral limb where postural video, maybe because it was smaller, maybe because it wasn’t being paid attention to, was not heated when the legs started to creep in. And if you look at you kind of the position from the central knee cushion to here versus this one, you can kind of see as it gets closer to that central area that it’s getting more and more dose. This is something that we’re now working on, make more robust as far as planning maybe a little bit wider avoidance sectors, but also conveying that information, Hey, this plan is using a no entrance, no exit on this structure. You need to make sure that the structure that’s far from isocenter we’re not involved in the treatment is also in the right position before and during treatment we need to do that. We know the ipsilateral leg was in the field correctly and that it was aligned appropriately because had it been moved out and violated the SGRT tolerances, the beam would’ve been turned off and that there was no indication of any gating during any of these deliveries.

Michael Tallhamer (20:28):

Here’s just a prospective avoidance of previously treated areas. This is another rendering that we can do via some Python code of the surface of the patient with previously treated upper lip and left cheek. We’re now treating the bridge of nose and nasal sept with a custom bolus. And we really, because we took these to 54 gray, are really concerned more of this region leaking down into the lip that was treated about six to nine months prior, making sure that this area, both that the bolus is in the right position, but also that the dose is not creeping into that previously treated area given the very tight margins that we’re dealing with. So very easy to provide these renderings to the therapist part of their setup images that show up right on the TrueBeam when they’re setting them up, and then also are available to them on the console while they’re delivering. Watching the DoseRT system monitored the dose for delivery. This one’s for the physicist in the room.

Michael Tallhamer (21:18):

This is a retreat head and neck that previously went to seven gray head floor of mouth recurrence. This was turf dust by the university. Very high modulation factor for these fields trying to avoid the parotid in the oral cavity that had been taken to previous tolerance dose. And so with the geometry issues in the head of the machine, we wanted to make sure that we didn’t see scattered dose underneath the wide jaws because the high modulation factor. So we just essentially ran the plan prior to physician review on Cherenkov emitting plate so that we can make sure that there was no dose due to the high modulation factor. It passed that we had the physician review, it ran a standard QA, and then we were able during delivery to three dimensionally verify that those avoidance sectors were not being violated because of these parotids. And the oral cavity had been taken essentially to full dose with the initial 70 grade treatment. So now we have multiple stages of QA in the plan prior to the physician review after the physician to make sure the dose is correct and then on the patient from day to day to make sure that the delivery is what we had intended.

Michael Tallhamer (22:18):

This is a planning technique valuation. We saw this breast boost that looked like we had some dose creeping over in the contralateral breast. Doesn’t look like a lot, but this is an arc-based treatment that is fairly well-loved by our physicians. And so we do this because it’s their preference over photon-electron mixed boosts. And so we started looking at this from not just a composite dose, but what is going on during delivery. So we can go back to these videos and look at the first field, nothing of note, the arc starts coming over the top of the patient. Everything looks relatively good, but what we shouldn’t see is what we’re about to see. And so here we see this dose tracking across the contralateral breast. And again, an avoidance sector that was not opened up far enough for the projection of the cavity resulted in the dose turning on long before it had cleared the contralateral breast.

Michael Tallhamer (23:04):

This was not seen in the planning system. This is another dose interpretation thing. This was not seen because of the rendering issues with Eclipse. If you change this, you lose some detail, but if you change this to a fractional dose rendering rather than a complete dose rendering, so the total dose, you can actually see these low dose regions that are across the contralateral breast and should have been picked up during a physics check. So taught us that there was some limitations within our treatment planning system as well and in interpreting the imaging that they were giving us.

Michael Tallhamer (23:34):

And finally, data interpretation. This is a breath-hold VMAT delivery breast that presented at our celebration facility with an unexplained hole in the Cherenkov signal. We turned this over after much deliberation to Cherenkov imaging consortium users as well, to kind of give us an idea of what could be causing this. They had the CT, they recommended that she has an expander and thought maybe the port was causing this, but the port is not in that plane. And so we went back and forth. SGRT tolerances tended to make the hole maybe slightly smaller, a different shape. And so it came down to a thresholding thing. The hypothesis was the thresholding of the dose signal because it was a low dose area that it was just taking it out. And we just was, we weren’t seeing it, unfortunately.

Michael Tallhamer (24:16):

A couple months later, another patient presented this time with no expander and another hole inexplicably that changed size slightly from day to day. This was sent from celebration to Colorado for like a secondary physics review because it’s a new technology. They wanted someone else to review it. I was reviewing the actual treatment plan instead of just Cherenkov imaging and found that the plan did not use flash. Unfortunately, it wasn’t using flash, and we always used flash, so that seemed odd to me. And then on top of that, the SGRT tolerances were three millimeters. So the hypothesis was that we were allowing the chest wall to excurs to exceed the field edge, and that’s what was producing this hole.

New Speaker (24:54):

So to test this hypothesis, we created an original plan. The original plan was already done. We created a plan with flash, and then we also used TLDs in the indicated hole from the Cherenkov imaging to see if there was a decrease in dose from the expected. Those TLDs came back while we were doing this plan in just 24 hours, showing that it was a 30 to 40 centigrade per fraction, or a seven to 11 grade reduction in dose in that region of the field due to the fact that the edge of the field was, or the chest wall was going beyond the edge of the field.

Michael Tallhamer (25:24):

And so we immediately switched over the flash plan, reran TLDs in that same region, and the TLDs came back within one to 2% of the expected dose. And so a slight misinterpretation of the shrink data on the first patient resulted in a second patient with a similar problem, which really required us to reevaluate and actually get a second set of eyes, which is why we work so closely with our sister site in Florida to make sure that we’re interpreting this data because it is new to all of us correctly and with the right intent.

Michael Tallhamer (25:52):

This study is out for publication now, showing the utility of Cherenkov imaging in rapid de-escalation of dose and escalation of dose from the Cherenkov imaging being identified. This is another one that should be going to publication soon.

Michael Tallhamer (26:09):

This was presented to me because of that poll in the dose we have a small breasted patient with a fairly large seroma. I was called to the machine because of a cold spot in the dose distribution. And that cold spot you can see right here, that cold spot was feared to be a hole because again, left-sided breast patient DIBH. But upon investigating a number of our similar breast patients, all these are very young breast patients. These seromas are visible in the images themselves. And so the composite images, we typically, if we have a very, very large seroma, we will evaluate those seromas via a CT to see if we need to replan these for our boosts. That does unfortunately require a young patient, oftentimes in their thirties to receive another chest CT to get a new plan. We’re looking at the utility of the Cherenkov imaging to actually flag us for seroma changes rather than using a cone beam CT or a diagnostic CT through the entire chest cavity and contralateral breast for these young patients to potentially flag us that maybe that’s needed before we just proactively give it to them to see if it’s needed. And so, maybe an interesting application, probably not the primary application, but something that for us at least should be investigated at some point because we can actually detect changes in these seromas from these images.

Michael Tallhamer (27:31):

So with that in summary for AdventHealth Cherenkov imaging in addition to SGRT is really clinically improving our quality and safety. It allows us to really rapidly address issues that we’re having with any of our patients, whether it’s a compliance or body habitus issue or just a plan robustness issue, something we didn’t account for during our planning process. It provides us a way to detect unexpected straight doses things like that, VMAT that was beyond the edge of the CT, and then provides a way to evaluate plan robustness. So we’re now really actively looking at how do we communicate our planning techniques to our therapist to make sure that the appropriateness of the setup matches the assumptions made during these planning processes. And then the Cherenkov imaging can be used to recommend and aiding treatment plan adjustments depending on if the field edge needs to be dropped. If we maybe have a chin or a compliance issue, maybe we need to open or shrink down a field. And then the Cherenkov imaging data itself while it can improve your quality of treatment and delivery, it does need to be heated in the data interpretation. The data is there, the interpretation is kind of where I think we as a physics group probably need to provide some guidance and help people with that interpretation into the future.

Ella Horgan
Radiation Therapist
Prince of Wales, Sydney, Australia

Video Transcript

Ella Horgan (00:04):

Hi everyone. My name is Ella. I’m a radiation therapist working in the Prince of Wales Hospital in Sydney. Today, I’ll discuss with you a case study titled Adapting to Patients’ Limitations. This case study discusses a head and neck cancer patient with multiple limitations, making his treatment a little bit challenging. However, with the use of SGRT along with daily Cone Beam CTs, we were able to treat this man completely massless, all while making very minimal moves with our imaging scans.

Ella Horgan (00:33):

This patient is a 91-year-old man diagnosed with a preauricular metastatic squamous cell carcinoma on the left side. After careful consideration, the radiation oncologist determined the most appropriate treatment approach would be radiation therapy targeting the left parotid bed and upper neck. However, this case presented significant challenges due to the patient’s numerous comorbidities including chronic obstructive pulmonary disease, a high body mass index, and increased risk of falls due to instability and vertigo. These factors made the treatment plan more complex and required careful management. As you know, the standard immobilization for a head and neck cancer patient would be a five point serum plastic mask, a comfortable headrest elevating the chin and a knee block supporting the knees.

Ella Horgan (01:22):

Some departments even use an open face, head and neck mask. However, due to the patient’s comorbidities, a modified immobilization approach was necessary. Given the severe vertigo he experienced, he was unable to lie flat. As a result, the patient was positioned on an elevated breast board with a 20-degree tilt. A VAC bag was utilized to provide head support and a knee rest was placed under the patient’s knees for comfort. Micro port tape was applied as an optional aid in the setup, which the patient agreed to as it had provided additional stability and reassurance, ensuring that he remains still. An IMRT plan was developed for this patient with the Isocenter strategically positioned 10 centimeter inferior to the planning target volume to avoid the risk of collision. The dose constraints for the organs at risk were all within the specified tolerances outlined by the EVQ guidelines. A 0.5 CENTIMES meter, CTV-PTV margin was decided by the radiation oncologist. This margin size is in line with the recommended guidelines for standard head and neck treatments using a mask. A study by Eser Al looked at the setup and fractional motion of using SGRT for head and neck cancer patients. The results of this study found the massless setup with SGRT and Cone BEAM CT were just as accurate as treatment with a mask. SGRT showed that inaction motion was gradual during the treatment and the CTV-PTV margin correcting for the inaction motion was 1.7 millimeters per OSUs treatment. This study agrees with our 0.5 centimeter margin being sufficient for accurate PTV coverage.

Ella Horgan (03:09):

After the plan was exported to SGRT, a region of interest was created. This region of interest included the nose, cheeks, chin and neck. After the plan was exported to SGRT, a tolerance of 0.2 centimeters was set across the translations and two degrees per rotations. The setting for beam control was turned on for the treatment to ensure any movement greater than 0.2 would pause the treatment. The patient was prescribed 48 gram and 20 fractions at 2.4 gram per fraction. SGRT was used for the patient set up and during treatment, a Cone Beam CT was created and used daily to ensure accurate patient positioning. A dummy run was performed on the Linac prior to day one to assess if collisions would occur. The result of this determined that HEXA pod could not be utilized, therefore, only translational shifts could be applied during treatment for a patient set up. Translations and rotations on SGRT were corrected for. The video function was also used daily to ensure the patient’s head and shoulders were in the correct position. After the patient was positioned on day one, the gantry was rotated around the treatment couch with imaging panels extended to assess whether the gantry blocked the cameras potentially causing the SGRT to flicker. This dummy run showed us that the cameras were not obstructed.

Ella Horgan (04:48):

Once this was complete, a Cone Beam CT was required. Due to the Isocenter positioning. The superior part of the PTV was caught from our imaging scans. This is another challenge experienced during treatment. However, as you can see from the Cone Beam CT above, there were enough slices to make a good clinical judgment on the patient’s positioning. All translations were then applied. The imaging panels were put away to ensure the SGRT cameras were not blocked during treatment. A new reference capture was taken to monitor the new acquired position and the SGRT response button was turned on. The results from the 20 fractions are shown here on fraction one, three and four. The largest translational ships were 0.6, 0.8 and 0.6, respectively. As you can see from the line graphs, the shift was decreasing. The shifts are decreasing from fraction four onwards. These shifts are 0.5 centimeters and reasons for this could be consistent staff treating him daily, understanding his setup, and altering the region of interest as necessary after the first few fractions. However, with the help of SGRT for set up, these post-imaging shifts were minimal. While the patient’s treatment was ultimately successful, several challenges were encountered throughout the process. As discussed previously, the Isocenter positioning presented some difficulties in reviewing the Cone Beam CT as certain slices of the PTV were missing from the image. Furthermore, due to the inability to use the HEXA pod, rotational shifts could not be made, which added complexity to the treatment setup. Additionally, the patient faces challenges in getting on and off the treatment couch. This issue was overcome by the presence of additional staff members during treatment, ensuring the patient’s safety and comfort.

Ella Horgan (06:44):

After completing the patient’s treatment, it was interesting to evaluate the result of the Cone Beam CT scans and reflect on the advantages and drawbacks of massless treatment. The benefits of massless treatment include enhanced patient comfort, cost savings for the department, reduced time spent on mass creation, alleviated patient anxiety and greater flexibility and positioning. However, the drawbacks include the need for monitor closer monitoring of the patient, specific patient criteria for eligibility such as pediatric cases, and the potential for longer setup times. In conclusion, by using SGRT, we successfully provided effective treatment to a patient with numerous limitations, all while eliminating the need for thermoplastic mass for immobilization. Importantly, the CTV-PTV margin remained consistent between the massless approach and the traditional mass-based method, ensuring the treatment precision was maintained. This approach highlights the potential of SCRT and Cone Beam CTs as a reliable and effective alternative to conventional immobilization techniques, especially with patients with treatment limitations. Thank you.

Sunny Chan
Radiation Therapist
ICON Cancer Centre, Australia

Dave Parsons
Assistant Professor & Associate Director of the Medical Physics Residency Program
UT Southwestern Medical Center, USA

Video Transcript

Dave Parsons (00:03):

Okay, so good morning everybody. I’m Dave. Today, I’m going to talk about how we really changed our practice using MapRT, which sort of echoes that again.

Dave Parsons (00:17):

I’m going to structure the talk is I really want to set the stage to why we wanted something like MapRT. So I’m going to go through what UT Southwestern is as a department why non coplanar treatments were important to us, how we made them safe in the past, how we now do that with MapRT and how we implement that into our clinic. And then go through a couple examples of where it’s useful and then summarize.

Dave Parsons (00:46):

SoUT Southwestern is in Dallas, Texas. If you’re questioning where my accent is or disappointed that it’s not a Texas accent, I’m actually from Nova Scotia, Canada originally the nice gentleman at the front lobby when I asked for directions thought I was Scottish. And I was like, I don’t think it’s that strange of an accent. But yeah. So we’re actually celebrating our 20th year as a cancer center, well not cancer center, but a radiation department at UT Southwestern. And you could see we started with just treating over 1200 patients a year. But now in, well, last year in 2024, we’re treating just over 4,600 patients. And even though we have added linear accelerators to our department, the main driver of that growth has been going to more hypofractionated. So back when we started, we were doing about 22 fractions per course of treatment. Now we’re doing about 12 fractions per course of treatment on average. And you can really look at why that is. So being in the US, we do have a healthy conventional population with tangent, for example, or whole brain treatments. But a lot of that other stuff being IMRT and SBRT that drives the department where SBRT or SABR is the largest chunk of our patient population that gets treatment.

Dave Parsons (02:10):

Now, if you ask my boss which is Robert Timmerman, how you should plan those patients, he would say the following, when you’re doing a SABR treatment, you need to respect the tumor coverage and the dose compactness and how that’s, that really defines SABR. So when you’re thinking about analyzing those plans, you want to be a very conformal plan. So pad, conformity, index approach, stream one low d2cm, a low grade index, I’ll go over those on the next slide if you’re not familiar with them. And what you’re doing to achieve those is to have really many beams coming from many different directions to spread out that entrance and exit dose effectively trying to create an isotropic fall off where possible.

Dave Parsons (02:55):

So just to go over those d2cm. So if we have our PTV there in red, a two cm expansion around it in any direction is that green structure. And we’re just trying to limit the max dose going into that. And I am a physicist, so I need my one equation per talk, and it’s just going over the grad index, which is the volume receiving 50% of the RX divided by the volume receiving a 100% of the rx. And if you’re familiar with lung planning, you would see something like this for from the RTOG, which says, oh, for your PTV volumes, you expect to have a gradient index somewhere here. And then the max d2cm in this column. And which usually you can achieve if you’re doing non coplanar treatments.

Dave Parsons (03:38):

Mu-Han Lin, one of my colleagues, had this nice slide that sort of shows that process. So here she has a lung lesion with two coplanar arcs which, you know, met the constraints with a grad index of about 4.6 a d2cm of 50% of the rx. But you can make that better by taking one of that, those complainant arcs and changing it to be non coplanar, where the gradient index is the same, but you’re getting that d2cm to be lower. And then if you go completely non coplanar, in which case she had 10 3D conformal arcs. So not even modulating the fluence, she’s able to get that gradient index to be lower and the dose compactness to be lower as well. And you can really see that there in the cyan. We’ll try to put it on that screen too, where that’s really becoming more and more compact as you increase the number of non coplanar beams.

Dave Parsons (04:30):

So hopefully that’s set up the stage to why we want to do non coplanar treatments overall, almost 40% of our patients are SABR and from the top down we’re being directed to do non coplanar plans to give the best quality we can. So how do we make that safe?

Dave Parsons (04:50):

So we did what you would normally do for us, which was you set up the immobilization on the couch and you check the clearance. The therapist here are showing that. So we use the elect body frame still, and you can see they set it up to the lasers in the room put the isocenter on the frame coordinates, and then actually ran through every combination of coaching gantry and collimator and says, does it pass or does it fail? And if it fails, could we recommend something that would clear that would then go back to dosimetry to say, yes, you’re good, or, oh no, you need to revise your plan.

Dave Parsons (05:30):

Of course this adds time. So this plot here is showing the number of angle checks and then how long it took from when the dosimetrist requested it to when they got it back from the therapy team at the machine to say, yes, you’re good to go. If your plan, you can continue your planning. And most of the time we’re averaging that to be within one working day for a dosimetrist. So a mean time about 6.2 hours. However, in some cases, say the plan was submitted at the end of the day, the morning, those symmetry or therapists didn’t get to it because they had patients to treat and they didn’t really get to it until the following end of day. So now you’ve added that whole day of just waiting to know if that planner can proceed with their treatment plan they wanted to use.

Dave Parsons (06:16):

And in some cases for these 60 patients we looked at, it was almost 24 hours. So, and that would be roughly three clinical days of working time.

Dave Parsons (06:26):

Of course, you can always get the scenario where you have drastic failures in this, where the plan failed on multiple reasons. So this is the feedback one of the dosimetrists got. If you can’t read it, we’ll just make it really bold for you and say your ISO two posterior, your cone beam won’t clear the panel, hits the couch or hits the immobilization, a their plan, send it back and hopefully get the all clear. However, in some scenarios, even though they aren’t having any issues with the immobilization or the couch being an issue and two of your couch gantry combinations also collide with the immobilization. So you have massive changes to do in that you’re doing for your changing your actual isocenter in the plan. And if you’re that the dosimetrists you’re probably starting to panic now.

Dave Parsons (07:02):

And then they would have revise axis with clearance, they might say something like, you’re all clear, but you should really watch the elbow fraction one. And you could see that for this patient, the elbow actually goes quite above the immobilization and they know that it’s there, but they don’t know where it really is. They’re just guessing that it could be an issue. And we’ll actually come back to this patient later in the talk to show it actually was an issue fraction one.

Dave Parsons (07:39):

So of course is what we believe to be a solution for this. I know some of you are familiar with it, but for those in the audience, I’m just going to give a brief explanation of what it is.

Dave Parsons (07:53):

So it’s a surface guided technique that captures the patient’s surface at the time of simulation. So that’s this guy here. And then it combines it with a LIDAR scan of linear accelerators. So here’s one of our, the true beams and they combine those two to look at how that would, how the clearance would be for a prospective plan.

Dave Parsons (08:13):

So if you don’t have this in your clinic, you essentially get two cameras on either side of your CT and then that is used to acquire the patient’s surface for us. We also have SimRT in our vault, so that’s the third camera there. We really wanted the monitor to be in the room. So we really proposed that initially when they installed it. So we would have an extra monitor for SimRT and MapRT in our vault and then also in the console two. But you can use the same keyboard and most it’s to between the two systems.

Dave Parsons (08:47):

After you acquire a surface you would get this interface on the software. You say, I want to use this machine model with this surface capture. In this case it’s a head and neck patient with laryngeal cancer.

Dave Parsons (08:59):

And then once you select that, you can proceed to this interface, which shows you your imported DICOM parameters here. So we have our imaging fields, our arcs for treatment, and I don’t know if I can do that over there, but yes, your imaging fields and arc for treatment. And then here you can see what that looks like where you have gantry on the Y and couch on the X axis, in which case we’d have two non coplanar partial arcs and then one full coplanar arc for this patient. And then of course your collision zones, which would be areas that you’d ideally want to avoid. You also have this interactive window which you could call like a room’s eye view where you’re going through the treatment. Which you’re going through the treatment to see what that looks like as if it were to happen. And in which case you can see for these two partial arcs, everything is fine. We have no issues there at the clearance. This is using a d2cm safety buffer around the whole treatment. However, if you were to navigate into one of those collision zones, you can see what that would look like, where we would definitely be hitting that patient’s pelvis if we continued to keep going, which we ideally want to avoid. So that’s really why we thought found MapRT attractive. Cause it shows us what you don’t know, which is where the patient is relative to their mobilization.

Dave Parsons (10:28):

So how do we implement that in the clinic?

Dave Parsons (10:31):

So we actually got this back in August, 2022 which was one of the research versions installed at our center. And then there was a second one in Raigmore in Scotland. And then over that next year we validated it, gave feedback, had iterative updates, and then despite it not being approved by FDA in the us, we as a department said, we’ll take the liability. We want this to be clinical now. We don’t want to wait for more updates to go live. So we went live in October 23. We then installed our second system on January, 2024. This also added more models in that update that happened then. So you can see we have a Halcyon here, a TrueBeam with its imaging arms out a TrueBeam with a electron applicator and a versa. But you notice the pendants down in this Versa model. Originally that wasn’t, which was a source of constant, is it going to clear or not? So now they’ve actually put that pendant down, which hopefully elect that someday we’ll just move it to the end of the couch. And then just last month we installed our third system, which is on a PET CT that we just installed. So we, we definitely think it’s the, the future and we’re, we’re pretty happy to have it on our three simulation machines. Hopefully, someday we’ll go on an MR simulator, but we’ll see.

Dave Parsons (11:55):

So our first question was what is the accuracy? And this has work done by Siqiu Wang, who was a resident in our department at the time, but has since been joined us as faculty. And she said, well, let me map that surface. The interface of that collision space where she took one of the cube phantoms scanned it, went through the process with capturing it and said every point of that surface, I want to see how close that is to reality. So she then went to the vault, set it up and said, okay, how close can I get without, you know, causing the sweat to pour down my head as I get closer and closer to the machine and see if I have a constant couch, what would be the gantry angle to cause a collision? Or if I have constant gantry, how much would I have to move the couch to cause a collision? We didn’t actually want to collide, I didn’t want to explain to the department leadership why their linac was broken. So we got reasonably close, probably a millimeter in reality.

Dave Parsons (12:47):

And what she saw was if we have gantry angle here and coach angle here where red is the predicted MapRT collision and blue is the measured band, either how much you move gantry or how much you move coach for a given angle what that measured result is. And then that green band says, well, what if you expand the MapRT result by a degree? In which case we saw that all the measurements that she did was were within a degree of the prediction. So for us, we say that the accuracy, at least our install at our center was within a degree either for gantry or couch, which we’re pretty happy with for being roughly a millimeter from colliding the machine.

Dave Parsons (13:30):

How does that compare to what we were doing though? So that same 60 patients that we looked at earlier for seeing how long did it add to your planning timeline, we also went back and saw what was the success ratio. So 55 of those patients, no issues. They, the therapist did the angle check fraction one treatment went fine, but for five of those patients, fraction one didn’t go so fine, in which case either the mobilization had to be changed such as it was a prostate, you could move an arm to get out of the collision. Or in the case of a lung you aborted the fraction completely and said, sorry patient, we need to send you home and redo your plan because we’ve, we have a collision that wasn’t identified. And when we went back and looked at those same patients with the MapRT surface that we had, you could predict every collision that they saw and then when they did their plan update, that was also still true in MapRT as well. So we’re saying at least for our patient population that we treated and can compare to real data, it’s a hundred percent successful in identifying collision risk.

Dave Parsons (14:35):

So with that of course we said we went live. So for our true beams, which we have four in our clinic, this was the number of angle checks we were doing for essentially the last three to four years, going back to November, 2021. And you can see it goes up and down but the average is somewhere around 38 angle checks per month. Since going live with MapRT we saw that start to decrease, cause we’re getting more and more confidence. Of course we still have another CT scanner that was installed which was here and we saw it go back up cause when we installed that second system, we also updated the software. So it took a two week pause and said is the new update as good as the previous version, which it was. And then after having both systems live, you can see we sort of flatlined here with roughly three angle checks per month that we still are doing. Now you might be saying why still three?

Dave Parsons (15:30):

Well we have a pretty diverse clinic and some patients that either are sim for our MR-Linac, which is unity from elect or CyberKnife have a mobilization that interferes with the coach markers that are placed in the CT couch for acquiring a surface. So if you’ve not seen an LEC frame or elect Unity simulation, they have this annoying couch top that runs the whole length of the couch and just prohibits you from capturing the surface. Whereas CyberKnife they just had this black thick pad about three inches thick, that’s for patient comfort and the therapists weren’t moving that prior to treatment or during their sim to capture surface.

Dave Parsons (16:10):

We’ve tried to mitigate that going forward. So we’ve actually moved the coach markers down further so we can at least capture some of these patients. But ultimately for the elective coach top for the MR-Linac you’re just limited to the field of view of the camera not being sufficient. Even though it is quite large, it’s a very long couch top that elective side to use. For CyberKnife though, it’s just simply reeducating the therapist say you could fold that pad, do your capture and then if they do go from CyberKnife to another machine like TrueBeam, you can have that surface available. So we hopefully mitigate that going forward. But we’ll see that in the future data.

Dave Parsons (16:46):

So how has that changed our clinic? So prior to MapRT we were doing about 38 angle checks per month taking roughly about nine and a half hours of vault time per month just to do that. So effectively you’re, that’s saying one day a month that machine was just being used, nothing but to do angle checks. Ideally that machine should have been treating patients but over the month it’s being used to do something that we can now do at MapRT. In terms of planning that added just under 10 days of just waiting to know if you could use that plan that you want to use. After going live with MaRT we’re now down to three manual angle checks. Those are usually, like I said, the Unity couch top or CyberKnife, those patients already usually have a delay in their planning timeline because they’re going from an attempt being an MR-Linac or CyberKnife to a C-arm Linac. So RA going long but even still those are only adding about 40 minutes of vault time per month, and only increasing the planning time by a month or not a month. That’d be way too long. Just under a day. And waiting to know if that plan is clear to go.

Dave Parsons (17:54):

So we’re quite pleased there. I just wanna go through some examples now of how you could use this.

Dave Parsons (18:00):

So that patient I showed earlier with the elbow in their mobilization, this is her plan and you can see it is quite a posterior lesion in the lung and it was in the testing phase of MapRT. So we weren’t using it clinically yet, but the angle check everybody said you should watch the elbow but we don’t know. Well sure enough fraction of that definitely collided with the elbow. It’s fairly close to the target site so you couldn’t readjust the patient. So we had to cancel the patient, the dosimetrists re-plan it by just doing a simple drop of the couch by seven centimeter and then that shift the collision zone over. Of course this was not with MapRT but you can see how that changed the clearance space. But you could see how it would be readily identifiable in MapRT as this is definitely an issue that you don’t want to have occur on your treatment.

Dave Parsons (18:56):

How could we have done better? Well, since you can see this clearance space, you have this lovely little section with no collision when your isocenter is in the target and you can really make a much better plan using non coplanar planning for map with MapRT. So realistically, this plan should have been something like this where you have this quarter arc here at roughly negative five and then this partial arc here at positive 20 degree coach rotation.

Dave Parsons (19:25):

And you can say, well how much time would that add? That adds essentially no time, cause if you start with two partial arcs, like here, you can go in the MapRT and just say, okay, how much do I have to change these arcs to be able to use them? And you can see this one, it’s still relatively close to the elbow, so maybe we want to go a little higher. Whereas this one, it’s pretty much in the collision space, we’re still avoiding it, but maybe we want to adjust that one a little bit as well. But you’re talking on the order of seconds, so you’re probably taking more time to transfer the plan to MapRT than actually figuring out how much you want to adjust your plan with MapRT to know what your clearance base is. In the end, this would be what that plan would look like.

Dave Parsons (20:09):

And then here’s the pick comparison between the two plans. So this is what was treated. You could see paddock and formulate index is pretty good 0.9, ideal would like perfect would be one gradient index5.4. If you look at the 25% grad index, 28 and a lot of MU. But you can see with that non coplanar patients essentially the same conformity. The gradient index for 50% starting to go down, but the 25% is cut in half and that 25% is this purple line here. You can see where it’s really quite different when you’re treating it non coplanar compared to just as two x or coplanar arcs. And actually it’s a more efficient plan too, where you’re cutting that mu as well.

Dave Parsons (20:52):

If we repeat that over 20 more patients, you sort of see the same thing. It is statistically different in that the conformity is usually a little bettern on coplanar arcs, but clinically is that significant? Probably not. What is probably this low dose bath, it almost is always better with non coplanar arcs. Some cases no. But that’s usually when you’re starting to get to larger size lesions.

Dave Parsons (21:19):

And you can also look at that here where this is showing the percent of the RX dose as you move away from the PTV where at the high dose it’s roughly the same for coplanar and non coplanar arcs. But after about eight millimeters from the PTV surface, you start to see that diverge where your coplanar arcs are in blue and your non coplanar arcs are in this orange color where it almost always will give you a better plan if you do non coplanar treatments.

Dave Parsons (21:46):

Another interesting site is partial breasts. So we have a healthy partial breast program at UT Southwestern where you’re just treating the tumor bed, not the whole breast. And I think from, well pretty much any breast patient, the number one thing you’re trying to spare spares the heart. As we know from Darby paper for every mean gray increase in the D mean to the heart, you’re seeing about a 7% increased risk of major Corona events in that patient’s lifetime. And a more recent paper from that same group for lymphoma patients show that there’s potentially no threshold to where you would see an impact on their long-term health of the heart. So ideally for breast patients, we’re pushing that as low as we can possibly go.

Dave Parsons (22:29):

But some patients that’s really hard to do. So this is a partial breast patient. You can see our CTV and PTV there in orange and red around the tumor bed and it’s really close to the heart. It’s about a centimeter at closest distance from the heart to the tumor bed. And if you treat this, this with coplanar arcs, you can see most of that exit dose is going into our structure we’re trying to avoid.

Dave Parsons (22:55):

But it is a nice open collision space where you can really use that to your advantage, but it doesn’t pay respect to the geometry of the patient. So you can see if you instead use one of these vertex arcs that it will easily clear here, you can effectively if the video will play, avoid the heart at every possible angle of that arc, giving you a much better heart dose what we’re trying to get to, but we know from lung patients, LAD’s probably a good correlate to heart effects and you can decrease that mean dose as well by using non than the coplanar arc would.

Dave Parsons (23:22):

How that looks like when you compare the two plans, you can see this low-dose bath going into the heart, whereas the non coplanar plan really is avoiding the heart both on the, the axial plane. And you can see what those dose differences look like. So the mean dose is decreased by just about half, whereas then the V 1.5 gray, if you plan partial breast plans, you know, that’s usually the metric we look at is also decreased by about a third. LAD we don’t really have a number for coplanar plans. And interestingly, lung also decreases by using the non coplanar treatments. If you repeat this over 18 other patients, you can see that plan quality is sort of these guys here. Pad conform index is about the same ratio of 50% dose to a hundred percent dose in the breast is the same lung dose, approximately the same contralateral max dose of the breast about the same. But all of these heart and LAD metrics for the D mean here, the V 1.5 gray and interestingly the V seven gray are all decreased. And similarly for the LADD mean is also to decreased, but what isn’t decreased is the max dose of the heart and the LED that’s about the same in either scenario.

Dave Parsons (24:46):

So with that I’d like to summarize the talk and say MapRT is a novel SGRT solution for clearance mapping. It’s more accurate than our manual angle checks, as you saw by the patients that we did in that example. It greatly reduces the planning time going from 10 days of wasted clinical time to just under an hour. It can give you better plans when you use for non coplanar treatments, which allows it to be effective and efficient in terms of your choosing those, those angles for coplanar treatments or non coplanar treatments.

Kitwadee Saksornchai
Radiation Oncologist
Division of Radiation Oncology, Department of Radiology, King Chulalongkorn Memorial Hospital, The Thai Red Cross Society, Thailand

Michael Tallhamer
Chief of Radiation Physics
AdventHealth Parker, USA

Ms Hajarrul Nisha Amjath Hussain
Radiation Therapist
Sunway Medical Centre, Malaysia

Cara Anticevic
Deputy Radiation Therapist
Peter McCallum Cancer Centre Moorabbin, Australia

Jeremy Hughes
Medical Physicist
Peter MacCallum Cancer Centre, Australia

Video Transcript

Jeremy Hughes (00:03):

So today I’ll be talking about CTV-to-PTV margins – very exciting. I know by basically looking at patient motion during treatment, I originally gave this talk for a more physics audience, so I do apologize in advance. I’ll cover a lot of the fun equations. How about that? We’ve got an institution agreement with Vision RT. I’m related to this. So let’s start with the basics. I’m sure you are well aware. We have a GTV that we want to hit. We put a small margin on that. That’s our CTV which is taking into account a lot of the microscopic tumor cells that we want to hit. We then do another expansion to the PTV. This takes into account a lot of the uncertainties, you know, machine geometry uncertainties, treatment planning uncertainties, just so that we’re confident that we are delivering the dose to the CTV.

Jeremy Hughes (01:03):

In intercranial SRS cases, that’s right in the middle of your brain, you want to be sure that you are delivering the dose to the tumor, but you don’t want to be delivering too much because you don’t want to be delivering dose to healthy brain tissue. So yeah, so in these cases, you want your CTV-to-PTV margin to be basically as tight as you are comfortable with. So when I made this talk our SRS that we deliver on our TrueBeam Linux with onboard imaging we used a margin of one and a half mils. So the question is, can we safely reduce this number?

Jeremy Hughes (02:03):

This is just a brief overview, we’ve got true beams. We’ve on board imaging typically three treatment arcs at catch 0, 45, 315, half a on the couch kicks. We do pre-beam imaging, a cone beam CT at couch zero, and we are still verifying with MVKV imaging at those couch kicks. We use AlignRT to monitor these patients with an open face mask with a one mil toll and a 0.7 degree toll.

Jeremy Hughes (02:34):

So I basically looked at a lot of our SRS patient data to see how they move. You’ve got all the real-time deltas being saved on the backend that you can pull out and all analyze. So I wrote a little script that does that. I also briefly looked at camera occlusion very briefly. It’s more of a sidebar than anything here.

Jeremy Hughes (02:58):

So this is a fun little equation. There’s lots of different ways to assess A CTV-to-PTV margin. This is just one of them. The main takeaway from this is that you’ve got systematic error and random error, and patient motion really falls into that random error component. There’s just some examples there for you. It should be noted that there is this equation that will get you a number, but in all of like, you know, the documentation and all the reports and everything they’re saying, you know, don’t take this number and just run with it. You know, this is in consultation with all the ROIs. There’s a lot of different factors in play here. However, the formula does not define an uncontestable correct margin to use clinically. And throughout this report, they restated that at least 10 times different ways.

Jeremy Hughes (03:58):

So we’ve got AlignRT out at Peter Mac, I’m in the Moorabbin campus, but we’ve got them all over Peter Mac. And it’s very basic. So there’s our little open face mask. We’ve got a ROI that we place just on that patient opening. You capture a reference image and then it basically monitors how much your live ROI is different to that reference image that you’ve captured. AlignRT is taking a lot of information in the backend. It does this every maybe 0.1 of a second or so. You’ve got all your delta patient movements, you’ve got percentage overlap there which basically is what percentage of your ROI is AlignRT using to do its calculation. So sometimes with camera blocking or something that will drop down. But you’ll still get the, the data.

Jeremy Hughes (05:02):

If you want, you can make some pretty graphs, and I love pretty graphs. This is basically a whole-patient treatment. There’s a lot of information. I don’t know why it’s automatically moving to the next one, but that’s fine. So you’ve got the whole patient treatment. There’s your three arcs – Orange is beam on, so there’s a first arc being delivered, second arc being delivered, third arc being delivered. And then you’ve also got your AlignRT thresholds here. And then, like reference images being taken for AlignRT. And you can see how the patient sort of moves in reference to that reference image over the course of the treatment. So zooming in on one arc delivery, you can see in this case, you’ve got vertical longitudinal lateral components, and then there’s your rotation pitch and roll. There’s your magnitude error as well. And I’ve just pulled out the couch rotation. So this is for couch zero and your little percentage overlap, which I’ve sort of used as a surrogate for camera occlusion. There’s a lot more to do in this, but basically, you can say, Hey, the ROI, there’s less ROI being used here in this little region. And you can kind of see it corresponds with a jump in the parameters of the patient motion.

Jeremy Hughes (06:29):

So there’s a lot of math’s that you can go into which I want. But basically I looked at our patient cohort. I did a lot of medians. I did a lot of standard deviations of how much they’re moving inside the actual beam itself. And you can kind of come with a systematic error of your patient cohort being around, Hey, let’s 0.1 degree and then your standard deviation, that’s your spread as well. So that’s only analyzing when the beam is turned on. And so it’s ignoring all those times where the patient has triggered an Interlochen AlignRT.

Jeremy Hughes (07:16):

I basically also looked at the percentage overlap as well. So I put thresholds to the data. So going back here, I looked at it and I basically said, all right, let’s analyze all of the data. Let’s analyze only when the data is say, above 70%. So that would ignore that data as well, just to see how the numbers changed basically. And all in all, it’s sitting at around, you know, 0.1, roughly point 0.1 mil a little bit more in the long, which is you, I feel like you sort of expect just from looking at a delivery you can see the long is kicking off a bit more.

Jeremy Hughes (08:04):

In these margin calculations, of course there’s a lot of caveats as there always is. In this one, I’m purely looking at the ISO being in the middle of a target. I’m not looking at lots of different targets with one iso. So, not getting into some of that funky rotational stuff that Michael was talking about in the previous talk. And there’s a whole lot of numbers, as he was saying, linear algebra, add them up in a quad chart, do some analysis and that helps you sort of come to a margin at the end of it. So this is looking at, I guess, more of a geometric margin. So really looking at that the geometric misses in theory, again, like I said at the beginning, there’s a lot that goes into the actual conversation about what a CTV-to-PTV margin is. But that definitely, I guess, put our mind at ease, you know, in talking to the ROS who are very keen on reducing that margin down to be like, well, one mil isn’t so far out from our calculated margin with all of the tolerances and assumptions inbuilt into it. So all of that to say we looked at a lot of our patient cohort and did our math’s as you saw before. And it really helped us to, I guess, be like, well, one mill for these cases is well acceptable, looking at our data. And camera occlusion didn’t really have a lot of impacts in the, I guess, patient information as a whole. But I think that’s another scope that you, we want to investigate further. I will say, just because I forgot to mention it before, all the real-time delta analysis is quite nice as well. You do have to extract it out from the machine, which I found was the, I guess, the biggest barrier for entry for just my analysis, just because you do have to like physically go down there and like click buttons. I tried to automate it, but because of that, I guess we’ve treated a lot of patients here, but I only got to analyze, you know, a subset of them basically. But yeah ongoing, this could be an ongoing thing. You could in theory, extend this out to all your patients treated with surface guided to, you know, really retrospectively just analyse the patient motion. But it is the case that you still need to physically go and export it, which I think is the main limiting factor of this. Yeah. Thank you.