Improving Plan Quality & Safety Using Surface Guided Planning and Dose Visualization

Adi Robinson, PhD, DABR (00:02):

I’m from AdventHealth Celebration. Here are my disclosures. AdventHealth Celebration has a CE agreement with Vision RT and some of the data used from AdventHealth Parker. They have the kind of agreement. So, a little bit about AdventHealth as a whole, if you haven’t heard it by now.

Adi Robinson, PhD, DABR (00:21):

We have multiple sites across the nation in about nine states. But the two sites that we gather a lot of data from is site here in Celebration in Florida. That has SimRT, AlignRT, MapRT, DoesRT and PatientID denoted by the little Mickey ears because we are at the foothills of Disney World, while AdventHealth Parker, which is here, has the same kind of line up at the foothills of the Rocky Mountains. If you wondered what our clinic looks like this, and then the Boring Parker Clinic looks like that you can judge for yourself what looks better. I’m not going to get angry, but we both utilize SGRT in every step from sim all the way to treat. And we, from that, we gather a lot of information and implementation processes about using SGRT for all these kind of modalities. So I’m kind of going to focus today on the planning phase and the treatment phase more focusing on surface guided planning and dose visualization.

Adi Robinson, PhD, DABR (01:37):

So let’s start with surface-guided planning. With surface guided planning basically from day one you can ensure that any plan that you have has a safe delivery and you basically can reduce to zero your need of physical collision checks. That’s based on using the surface guided clearance map and putting fields on in the allowed non collision areas. And with that baseline, without mastering that, you can move into coplanar optimization where you can start picking your treatment fields based on allowed location and start tweaking them to kind of improve your dosimetry. You can do that to improve target coverage as in reducing OAR dose, just like the talk before me highlighted, and I don’t need to do a lot of convincing that non-coplanar works because, again, previous talkers did it way better than me. But that’s kind of the, the final step, the introduction of non-coplanar planning with surface guided planning to basically get to a point where you’re creating the best and safest treatment plan per patient.

Adi Robinson, PhD, DABR (02:45):

So a little bit about our setup. So this is our sim room. It’s loaded with SimRT and MapRT together. I believe this is the smallest sim room that they were able to put the cameras in and we had to even remove a cabinet to make it fit, but we make it work. Just to test it, we have a, a therapist that is about 6.4and he’s our Guinea pig. We kind of made sure that we can image him, and he’s basically the length of the table, so we can grab a picture of him from head to toe. So if we can do him, we can do 95% of people. So these are the MapRT cameras, two lateral cameras and a workstation inside the room and outside the room. And just for reference, these are the SimRT set up and unfortunately our control area is really starving because of all these monitors and all that stuff. So kind of a plan ahead if you ever get both products.

Adi Robinson, PhD, DABR (03:51):

And this is what a clearance map would look like. I know it’s a lot of it’s kind of busy, but you can divide it into basically kind off our or five sections in the middle. You’ll see the actual clearance map with collision areas and allowed areas. The x-axis is the couch, the y-axis is the can tree. Fields that are arcs are going to be shown with a line, a solid line with a dot delineating where the arc is going to start. Just a little yellow dot is a static field. Over to the left, you’ll see the imported plan if you choose to import it in, and you’ll get different signage of whether the plan clears green collides red, yellow means a potential transition collision between two fields. All the way to the right. You’ll see the virtual Linac room, which will be customized to your current setup. So if using a Varian versus an Elekta versus InBore system, even couch tops and accessories, that all can be customised. Above that is your safety buffers, which you will set for couch buffers and patient buffers. And to the left of that, you have your isocenter location and your couch shift. So, you know, you can, everything you need into in planning, you can find here. So you can pick the correct fields, you can change the isocenter and kind of see if that clears or not. You can even change the indexing, ship the couch if you want to test that. And that’s basically all you need to kind of optimize your plan.

Adi Robinson, PhD, DABR (05:23):

So our kind of workflow that we’ve kind of come up with is in our sim room, we try to capture a CT sim before we even, sorry, could I capture a surface before we even CT sim and kind of do a rough check of collision. My lead therapist is here also and she’s giving a talk tomorrow so she can kind of glance over this. But we are able to adjust positioning or immobilization device accordingly before we even initiate the first scan. We can even choose whether, if we think the patient’s not going to clear prone, we can change them to supine and try that and make all these kind of adjustments. Then the planner gets the DICOM information plus the clearance map so they can start optimizing the plan accordingly. And then the therapist at the end of the planning phase get the clearance report knowing that the plan is safe to deliver, and then the plan itself.

Adi Robinson, PhD, DABR (06:23):

So let’s look at a few case studies in different ways that we’ve looked at it. And, you know, the simplest cases you got a breast case for example and you just want to make sure that it clears your normal way of treating.

Adi Robinson, PhD, DABR (06:36):

So you export your final plan into MapRT and you’ll get something like this. You’ll kind of hone in on that field that’s really, really close to your collision zone, and you’ll know, lo and behold, no surprise, it’s the elbow. And you’ll see that you’re far away enough and you really don’t need to make any changes. This is good. You can print that, send it to physics and a therapist and kind of take a look.

Adi Robinson, PhD, DABR (07:00):

Well, what if you want to tweak some of our cases? And this is something I’ve learned from UTSW because they kind of showed me a lot of APIs and their ways of doing it, and I kind of wanted to try it that way. So here’s one A PBI that kind of came across and I wanted to see if we can treat it not a coplanar, and how would that look like?

Adi Robinson, PhD, DABR (07:20):

So here is the regular clearance map with coplanar plan. It shows a collision of one of the imaging dummy imaging fields that we put, so we can ignore that. But this is kind of our standard constraint planning constraints that we use. And everything looks good, you know, no complaints. But again, this is the training exercise. I wanted to work with our twos symmetry to see how that’s going to work. So let’s try that non coplanar.

Adi Robinson, PhD, DABR (07:48):

So I kicked the couch by about 15 degrees. Reoptimize, tweaked it a little bit, spent no more than about half an hour on it. And this is what I got. So the hotspot is about the samea nd the hard dose, the mean hard dose is about the same, but I got a good reduction in that long dose by about almost 10%, 9%. So that is kind of what we want to see. And that kind of enters the room. The role of like ARIA like a plan that clears all of your objectives is great, but if you can do better you know, ideally we want a hundred percent of the dose to the target, zero dose to OAR that’s unachievable. But if we can get them as low as possible with just minor adjustments to our planning, that would be amazing. And that’s kind of what we use this tool for. But this means nothing if those non-coplanar angles don’t clear.

Adi Robinson, PhD, DABR (08:41):

So we always have to check those. And by checking those, we see those two arcs clear beautifully. So no problems there.

Adi Robinson, PhD, DABR (08:51):

And this is a case that came recently. It was a left deltoid case that I was really, really, she had previous irradiation to the chest wall there. And we really, really wanted to get the OAR dose as low as reasonably possible.

Adi Robinson, PhD, DABR (09:08):

So again, we kind of compared our standard coplanar plan and our non coplanar plan, and we can clearly see that our non coplanar plan achieves a much lower hotspot dose in our or, and better coverage. So again, we got to really make sure that it clears because it was a really weird setup. And it kind of, it looks great. So we’re really, really pushing it with that gantry really close to the couch. But it clears. And with our physics commissioning and validation, we know that we can get it as tight as it, the clearance map says, and we know there’s not going to be collision. So these are, are some clinical examples of how we kind of implement surface guided planning. The way everything works with your treatment planning system is, for example, if you have RayStation, there’s full integration with MapRT and you can see it as a window within your treatment planning system. So, as you keep reiterating, the plan you’ll get updates to that clearance map, and there is a tab down there. If my laser pointer works, down there you’ll see like green check marks and then an orange one that basically tells you that the field is failing. And if we kind of update that field, it will pass. So with RayStation, you get your full integration and if you have Eclipse, just hang on tight. Mike will give a talk later on about the API and do a little bit of a show and tell.

Adi Robinson, PhD, DABR (10:42):

So that’s I’m, publishing your talk mic. So with that, we can switch to surface guided treatment with dose visualization.

Adi Robinson, PhD, DABR (10:56):

And what is that all about? You’re basically getting real time dose visualization of dose during the delivery. On top of patient positioning, you can really help prevent errors in real time and improve your clinical outcome.

Adi Robinson, PhD, DABR (11:10):

Where does this all come from? Well, it comes from Cherenkov Radiation and that Cherenkov Radiation is emitted when a charge particle moves through a medium faster than the phase velocity of light in that medium. And if you watch a little video, that’s a micro nuclear reactor being turned on in water, so that blue light that’s emanating is the ran off radiation. And it’s a real effect that you can see with your own eyes. It was first observed by Pavo Tran off in 1934when he saw blue light, when he put a radioactive source in water. And then Tam and Frank kind of developed the theory behind it later on, and they all got the Nobel Prize in 1958.

Adi Robinson, PhD, DABR (11:52):

So taking that and looking at KO imaging where that shrink of light can be seen on the patient’s skin surface during treatment with special light sensitive cameras. And that’s a result of the interaction of the entrance and exit dose during your treatment. It’s a bit tricky to get it all working because, you know, we have kind of a ceiling heavy situation now in the Linac where you have the two you have the AlignRT cameras in there, and now you have two Cherenkov sensitive cameras in there and you got to make it all work so they don’t block each other and kind of saturate each other. So it’s a really, really cool set of engineering. But at the end of the day, that allows you to visualize the radiation treatment directly on the patient’s skin, which is in real time, which is pretty amazing. So it would look something like that. That’s what the therapist would see and that rank of light can be seen on the patient’s skin surface.

Adi Robinson, PhD, DABR (12:53):

So here’s one of the first cases that we have observed in our clinic. It is a, a bolus misplacement that happened on one of our patients. And so you could see the video as it’s running. We’re treating a super clave right now, and it’s going to switch to the other field. And you can clearly see there’s an area that is not covered by the bolus. So you are basically over dosing that area. And then there’s an area towards the media part of the, the medial part of the patient that is not getting covered on the right, you would see the two cumulative images that result from the end of the fraction. The top one is the result of the fraction with the misplacement of the bolus. The bottom one is the correct visualization of the bolus, and you can see the differences between the two. And that is quickly something that can be corrected and fixed for future delivery.

Adi Robinson, PhD, DABR (13:49):

Here’s another example that we’ve seen. We’ve seen recently an uptake of this contralateral breast dose. And you can see here in the delivery where there’s some signal there now, but that’s not the concerning part. This is the concerning part where we see a lot of activity in the contralateral breast. To the right of that, you’ll see the AlignRT it’s DIBH treatment, and you can see that the AlignRT thresholds were all met, so we were thinking we’re doing the right treatment, but and planning wise, there was nothing kind of out of the ordinary. And you can see the cumulative image down there. There’s definitely some signal in the contralateral breast. So that kind of, we, again, we were fairly new to the process back then, so we were trying, we said, what if we adjusted the threshold? We made it tighter so she could have a more consistent and tighter breath, and let’s see if that helps. If that didn’t work, we would have to either take the breast or replant it. But here is the result of the next fraction where we tighten the thresholds a little bit more, and you can see there’s no signal there. So that’s definitely a positioning breath hold kind of situation where even with our, you know, three millimeter plus minus three millimeter breath hold tolerance for DIBH, there’s still some play in that that resulted in some contralateral dose that is now prevented.

Adi Robinson, PhD, DABR (15:16):

Here’s a recent case of dose to the chin, and you can see the, the cumulative right there. And here’s the delivery of the video. Luckily, these things can easily be solved by firmly telling the patient to look more towards the other side or look more upward away from the treatment, and it should kick in about the next field when we treat the super cloud. A lot more activity. And then when a therapist sees these kind of things, they flag it and they make physics review it later, then we look at the cumulative and see what we can do about it. If the therapists think that they can adjust the positioning live, they have the ability to do so if they choose to.

Adi Robinson, PhD, DABR (16:07):

Here is a prostate case where the patient just lowered their hand into the field and you can see the right most images, the video, and you can see the doses going through his hands. He was just, he was supposed to put his hand all the way up to the chest, and unfortunately, it wasn’t noticed until the next field where it was corrected. That was, that wasn’t great, but luckily fixed. And here’s case of an extremity. Extremity is always kind of interesting because it’s, sometimes it’s a difficult setup, but with AlignRT and Postural Video, we really, really perfected that setup. But it was really, really cool to visualize it with DoseRT because we were able to kind of see the treatment being happening on the knee that we’re trying to treat versus getting any dose on the other knee and the other healthy limbs. And you can visualize that in real time, which is really cool. So you can see the entrance and exit dose right there and you can see the cumulative in the bottom, and that was, that was really cool.

Adi Robinson, PhD, DABR (17:17):

And here is an SRS an example, and I kind of isolated one lesion just for clarity, but it’s a vmat example and you can see a little bit of exit dose there, but as the dose rate will increase, you’ll see the high activity of treating that, and then you can see kind of the cumulative to the side and a few kind of screenshots of the other, of the exit and entrance dose from the two sets of cameras.

Adi Robinson, PhD, DABR (17:47):

This is another interesting case that I observed is Thymoma vmat plan. With vmat, you kind of see that spread, that’s that band, but you’ll see the most activity where the lesion is located. I just find it interesting becausewe you clearly see those two bands, and that’s split by the MLC. And I kind of went over and I tried to find that kind of location in the plan where the MLC is like that, and you could find it. So it’s, it’s right there. So it’s really cool to match the MLC view to the terrain coffee view. And we play that game a lot in the Linac. So it’s really interesting to see.

Adi Robinson, PhD, DABR (18:28):

And, you know, based on some of Mike’s initial work, you know, you can, you can do a lot of, you know, physics tests to kind of challenge yourself and prove to yourself some other kind of factors about that. Your cough signal and like the first thing that everybody wants to know, how is it, is it linear with dose? And can you calibrate your system to be more of a qualitative test? And I think we found that this would be the first test that you would do. You would check whether the signal is kind of generated linearly and it does if you just do a simple kind of step plan where you go from 25 centigrade all the way to 300, and you measure the mean linearity in itand then you plot it and you can see you can make fancy plots and make it look pretty, but at the end of the day, it’s linear, which is promising. But we wanted to take it a little further and kind of check different things with the system because we are going to be using it every day to make assessment and judgments on treatment plans.

Adi Robinson, PhD, DABR (19:25):

So we wanted to see, for example, whether or not the signal is consistent. So we measured the mean signal intensity on all photon energies for a period of three months every day and kind of extracted that mean signal out of it, and it was pretty consistent. It was plus minus 6% variation from mean, I think we could have gotten it a little bit tighter, but we’ve had some room illumination issue that kind of affected some of the outlier dose. So I really think it’s the real number is around 4%, which is pretty awesome and pretty reliable. Something else that we wanted to test is that, because we’re making a lot of visual judgments on the beam and the beam shape, I was kind of curious to see if it kind of holds geometric consistency or constancy.

Adi Robinson, PhD, DABR (20:16):

So we kind of, I picked a 10 by 10 field and it doesn’t really look like 10 by 10 when you look at through the ran off cameras because of there’s some geometric distortions, but did the same thing for three months every day, kind of measured the penumbra and kind of measured the distance between the two and kind of made day one to be the reference and kind of compare to what throughout the three months. And I gathered no more, no bigger variation than 2%, which is kind of like TG142 for us physicists of field size constancy. So that’s really encouraging. So that’s really cool.

Adi Robinson, PhD, DABR (20:53):

And this is the wishful thinking slide because I would love to see a shrink of signal versus Linac output. I would love to see the signal change as the Linac output change. So maybe one day it could be another tool in our chest to do daily output checks or a quick, you know, output check. So, you know, on the top right you’ll see the Cherenkov off signal as it varied within three months. And then on the bottom, the output, which was measured at the same time. And you know, on the left is kind of, I kind of put a fudge factor in to kind of put them all in the same scale. As you know, the Cherenkov, the output of the Linac lives between the differences between the two cameras, the output of the two cameras, which is great. It kind of moves up when the output moves up and it kind of moves down when the output moves down, but not in the same trend as I would like it to be. So more work to be done there, but encouraging, but need some imagination.

Adi Robinson, PhD, DABR (21:48):

So in conclusion you know, MapRT provides a clearance map that eliminates the need for collision checks and dry runs, and assist in improving the quality of the treatment plan. And DoseRT will give you dose visual, I can’t say that word, visualization in real time and assist in improving the quality and safety of a treatment delivery.