SGRT Treatment: QA Commissioning of SGRT

This is going to be a talk for commissioning AlignRT as a multi-clinic experience. The results are going to be based from our more recent acceptance and move towards SGRT at Vanderbilt Medical Center. I say multi-clinic experience because we have our downtown clinic and then about four different satellite clinics, and we’re slowly rolling out AlignRT and other SGRT systems to each of those. So we started this about two and a half years ago, and so far, we are quite happy with the results. No disclosures to report, no financial interest.

Vanderbilt is a bit eclectic in our SGRT applications. We started with BrainLab ExacTrac and Varian RPM. Vision RT is somewhat new, like I said, two and a half years ago. And going forward, we have more in our back pocket coming live soon. So as far as Vision RT’s catalog is concerned, they are pretty well established in multiple different systems. Within our group, we have just commissioned the SimRT system, that just went live about a month ago, and the AlignRT system. So we are primarily focusing on motion management on the therapy side. Within that AlignRT, we are now live on our TrueBeam systems. We have three of those up with one more in the works, and our Ethos system. We had one of those go live last December, and then our most recent one went live yesterday. So, we’re excited to have a lot more data coming in on that one.

So what we’ll be covering. So in our case, commissioning for me on the physics side is going to start when the ink dries on your contracting. That’s when we take over for our pre-acceptance testing, and that acceptance will go through the design of your commissioning checklist. And then, we’ll talk about the summary of our results at Varian. Commissioning for us includes developing commissioning workflows for all the clinicians all the way along the way through treatment, and as well as designing routine QA systems. And then at the end, I’ll talk briefly about some of the things that were unexpected. So learn to expect the unexpected.

Pre-acceptance. Like I said, this is going to start when the ink dries on your contracting and end as you meet with the service engineer for vendor acceptance. So from the physics side of things, that can sometimes start with your site surveys. Sometimes it can go back a little bit further to vault design, depending on if this is a retrofit or an all-new system coming in. In an ideal world, you would be working with your Varian rep and service engineers while you’re designing the blueprints of your vault and installing a new linear accelerator. And I will talk about a little bit of that as we go along. But assuming that you’re doing a retrofit, and we’ve done both, and I would again highly recommend starting or installing both these systems at the same time. But a site survey is going to probably be your first introduction to Vision RT. You’re going to have service engineers and contractors coming through.

One of the first aspects they’re going to look for is camera pod placement. You want unobstructed view from the camera pod to your machine isocenter for a TrueBeam, or the setup isocenter for an Ethos system. Second one is going to be the in-vault computer location. That’s going to be where your therapists are using the AlignRT system for initial patient alignment. They like to do that in the room without having to go back and forth to the console. Third one is going to be the in-room lighting control. So if you’re using one of the slightly older Vision RT systems, it’s going to be using the CMOS cameras. Those are going to be a little bit more light sensitive. I think their new Horizon ones are a little bit less so. But we ended up having to set very specific room lighting controls whenever Vision RT is being used. Camera pod power switch. That’s going to be vital for your morning QA, and all your cameras are just going to be on a regular light switch. So you’re going to bring in engineers as well. Light switches are going to be a surprise to me. Light switches and wall space are going to be prime real estate in any kind of vault setting.

UPS backup. So if you have any intentions of having the system stay up online if there’s any kind of power flicker, especially if there’s any kind of billing or documentation used by your Vision RT system where you want to put this on some kind of uninterrupted system. And then finally is going to be the IT infrastructure. Installing teams may expect many of those pre-installation visits, and they’re going to start by looking above your ceiling tiles.

So like I was talking before, camera pod placement, it’s going to be very big. Camera pods are going to be at a fixed location for most of these systems. In a standard TrueBeam system, you’re going to have your three pods, one on each side of the patient and then one directly above the end of the couch. Those should be in a place that are not likely to be bumped, not likely to be vibrated. You probably think my system’s not next to a train station, so probably not too much vibration, but we have found camera bumps are not as uncommon as I would like. So we ended up moving a lot of the other storage that we keep in the vault, patient backlogs, any kind of masks that might be up on a higher shelf. Those have been moved away from our camera locations to avoid those issues. Second one is going to be thermal sensitivity of these cameras. Again, as we move to the Horizon cameras, we’ve found them to be much less thermally sensitive. But, if anyone has a standard LINAC, and every LINAC vault I’ve ever seen has these omnidirectional HVAC vents up in the ceiling. These tend to blow out across the ceiling, right over your cameras, and then down to cool and heat the room. These cameras are very thermally sensitive, and we found as AC systems kicked on and kicked off, our cameras were moving on the reproducibility. So, two options we came up with was redesigning our venting systems, which means taking them out of the ceiling, bringing in an HVAC team. A little bit more complicated. So we opted for a simpler approach, which we put in these redirects, which you can use to block airflow directly of your cameras. Very simple solution to a very common problem.

Next one’s going to be the in-vault interface. This is going to be the computer system that they’re going to be installing in the console room and in the vault. These are mirrored systems. This is going to be probably the most important facet for your therapy side. We talked earlier about the posterior alignment add-on to Vision RT. Our therapists have found that to be probably their favorite feature in the system. Very useful for extremities and breast alignments. So, when your therapists are looking to use an in-vault system, first of all, they want to either be able to see the display from next to the patient. So what we found there was a wall-mounted system with a larger monitor was much more conducive to that, if you had the wall space. The second issue was when they are working directly in front of the console, they don’t want to have to be looking over their shoulder to check the patient and then back at the console over and over again as they direct the patient. So having that wall-mounted system on a bracket that allows it to be rotated so the therapist can view both patient and console at the same time was another big benefit to us. We see the difference between the clinics that have that and don’t, basically reflected by the therapists that enjoy using the system and the ones that don’t. Like I said, unobstructed view of the patient from the computer.

Okay. Next one, pre-acceptance. As we’re getting closer to the acceptance date, verify your LINAC’s performance conforms to the clinical standards at the most stringent technique that SGRT is going to be used for. One issue that we came up with for one of our installations was, during acceptance testing, we were seeing some abnormal results. What we ended up finding was our laser alignment was actually off. Our physicists had done their last monthly QA about 30 days prior to that, and we had seen about a one and a half millimeter drift. And that ended up causing some delays and unnecessary uncertainty during commissioning. So go through your mechanical and your imaging monthly QAs. I recommend doing this the night before acceptance. Check your couch and gantry translational motions. Check your front pointers and your ODIs, and then check your radiation imaging isocenter coincidence. So break out your Winston-Lutz phantom just to have a little bit more confidence, especially if you’re going to be commissioning the system for SBRT and SRS.

Next one’s going to be communication with your installer. Reaching out to your installer ahead of time saved us a huge headache. It allowed our commissioning team to know exactly the checklist for patient preparation, for phantom preparation, that was going to be expected on the day of acceptance. Discuss the responsibilities and timelines with your installer and verify, this is a big one, verify network connectivity. Which for us, we ran into issues with read and write capability on the firewalled Vision RT system back to our hospital’s network drives. As these hospitals are becoming more and more locked down with IT security, this is one of our big hiccups with any kind of ancillary system. So in our particular case, we had to delay acceptance two times midway through due to network communications failures. So I recommend testing what you can and being in contact with your acceptance engineer ahead of time.

Now we move into the acceptance side. This is going to be brief. This is a very streamlined and structured process. Your engineer’s going to show up. You’re going to have about one evening procedure with them as a physicist, and they’re going to be testing this thing to the strict standards of your vendor spec. So this is not going to probably count sufficiently for acceptance of the system for the commissioning standards. So that’s going to be quick. Expect to just kind of follow through, paint by the numbers in that case. If you have everything set up, expect it to be done in just a couple of hours.

Commissioning checklist for us. As a physicist, we like to have our sources to go back to. TG 147 is kind of the gold standard that we set for our non-ionizing patient setup devices. 147 is a little bit dated. ESTRO and then TG 302 have come out much more recently with much more focus on the SGRT side of that. So 302 is what we use in our clinic. It gives very specific patient positioning and commissioning guidelines, as well as routine QA. The recommendations include and build on those in TG 147.

So, your commissioning objectives are going to be what you dictate as the physicist when you’re designing what you intend to use your SGRT system for. So the first thing is to determine your treatment techniques. Is it just going to be IMRT? Are you going to go up to SRS? Or, that should be SBRT. In most SGRT online clinics, beam hold and breath gating is probably going to be what you’re most excited and looking forward to. But there’s a few others that we had to run through. We had to establish camera performance. We had to look at the interface with the peripheral systems. We had to look at that spatial drift and camera reproducibility, static and dynamic localization as they apply to your clinic, and then you should always finish this with an end-to-end testing that’s applicable to the treatment sites that you’re going to be going live with.

So, in our clinic, the interface with the peripheral system started with verification of data transfer. AlignRT has a local repository for DICOMs. Those need to be exported from your treatment planning system. For us, that was Eclipse. So verification of data consistency from your DICOM sent over through your export filter should be checked. Data integrity should verify that any kind of data abnormalities in your DICOM set, such as patient positioning, head first supine versus feet first supine, and so on, all those systems are going to be working. So, those need to be checked with patient exports for any kind of patient orientation that you can imagine. Integration with TrueBeam. When a patient’s loaded on the TrueBeam system, in this treatment mode itself, AlignRT should automatically be in communication and should load that patient and plan from its database. I don’t think that this works with Ethos. I think it has to be done manually.

Interlocks. Authorization pending. That’s going to be one that anyone who uses this system should be looking out for. And if that doesn’t show up and you’re not using beam hold, then something is going wrong. So when AlignRT uses any kind of motion management, it’s selected. Authorization pending should come up as an interlock as the system is waiting for verification. Patient is in breath hold before treating. Again, that interlock exists for protons over the ADI. Does not exist for Ethos. Functionality of the ADI for the TrueBeams needs to be validated. For us, beam hold capability is checked with a sort of de facto made in-house moving phantom. We set this on the RPM motor. It allows the leg phantom that came with the system to be used in a sinusoidal motion that simulates a patient in free breathe. If you can have something that’s more specific than this, it helps. But in the back you see our output verification, which is going to just be solid water with an ion chamber. So in that, we’re checking, A, the application of the ADI is pausing and allowing the beam delivery when appropriate. And then we’re verifying that dose is consistent. So we have dose measurements with and without beam pulse.

Next one is spatial drift and reproducibility. Spatial drift is going to be the test of the measured drift for any kind of stationary object as that camera warms up. In this case, we allow cameras to cool down as if they were coming on overnight, with a phantom cube set up to in-room positioning. So laser alignment for us. As our lasers warmed up, we measure the deltas in all degrees of freedom for phantom alignment over a 30-minute period. And in our case, we found that these cameras tended to take about 15 to 20 minutes to really stabilize. At that point, they basically were stable indefinitely. So in that case, we set a guideline for 20 minutes from camera on to any kind of morning QA or patient procedures. Static localization is going to be a positioning of some kind of phantom. In our case, we use the AlignRT cube phantom. And then applying a known shift and registering those shifts were within tolerance of the system.

So, some results from our clinic. We have their set-up cameras here, which is going to be from our TrueBeam system. This is going to be the three cameras. And when no cameras are obscured, we found alignment with one millimeter and five millimeter shifts to be well within tolerance, far below submillimeter. With one camera blocked, which is going to be realistic for any kind of IMRT applications, discrepancy did go up. That’s something to keep in mind, especially when you are designing your ROIs on the patient surface, as was talked about earlier. Now, I just want to jump into the in-bore side of things. Yeah, there we go. In-bore measurements were the ones that shifted from the set-up position to in the bore of our Ethos systems. In these, we found that camera to be a little bit more susceptible to large shifts. We found these to be about one, one and a half millimeters inaccurate, and that kind of segues into some end-to-end testing, which we’ll talk about in just a moment. Dynamic localization, not something that we really focused on. The spatial accuracy that’s relevant to breath hold was validated by our static localization test. But if you are using beam gating, you need to verify temporal accuracy of the systems and frame rate characterization.

Moving on to end-to-end testing. Overall system functionality and accuracy were evaluated by taking a test phantom from simulation all the way through treatment. For basic treatment fields and setup fields were applied to the simulated phantom with a radiation isocenter placed at a known target within the phantom itself. Each phantom was then taken into the vault, aligned using AlignRT, shifted to isocenter, and then that positioning was verified using onboard imaging. So in kV cone beam CT and then MV/kV planar imaging. So for our TrueBeam system, we found this very reproducible. Again, sub-millimeter. This gave us a lot of confidence that these systems could be used in a tattoo-less clinic without daily imaging, which gave us some confidence when we went live with our breast DIBH patients. Ethos, we ran into slightly larger margins or discrepancies. So as we shifted from our setup cameras to the in-bore cameras, again, we’re seeing higher discrepancies than we saw on the setup cameras alone. Not really covered here, but I think I recommend you talking to the vendors outside. The in-bore system is basically adding one more layer of complication. You are setting up to one camera isocenter, your setup field, and then you are applying a known couch shift to your in-bore. And our in-bore cameras don’t have three cameras. They’re basically just the one in the bore, in the ring. And due to this magnitude of error, which I think we are still looking to see if we can diagnose and remove, like I said, we just went live with this system, we’re still relying on onboard imaging for all of our patients. But that doesn’t disqualify it from being used for breath hold because that’s a relative set of patient motions.

So real quick, as a secondary part of commissioning is commissioning workflows. That’s going to be a multidisciplinary requirement in any clinic. I recommend you bring in your dosimetry therapy and oncology teams all the way through. From the planning perspective, it’s very straightforward. Simulation perspective is going to be a little bit more complicated. Your therapist is going to be tasked with coaching patients through breath holds. I always say the simulation day should be the hardest day in the patient’s radiotherapy treatment. If you are expecting to have between three and six breath holds over the course of a treatment field, you would want to have at least twice that number in the day of simulation. So we found practice breaths were very vital to this, going into multiple breath holds, and in cases of deep-seated tumors, multiple breath hold simulation scans. It allows us to look at reproducible breath hold and elevation of the chest wall and any other metrics that we’re looking for, for patient candidacy. Finally, we have a team huddle before the patient leaves any kind of simulation vault between the physicist, oncologist, and therapist, conferring whether there’s any clinical benefit to DIBH for these patients, ideally before sending the patient home. And those metrics are going to be defined clinic by clinic.

Treatment planning is going to be very straightforward. Once you determine this, it’s going to just be discerning your reference images, transferring those out of the system, and then importing into AlignRT. DICOM export. I always recommend automation when possible. DICOM exports from AlignRT can be set up directly. It’s a one-click process in Eclipse. It exports just your RT plan and your structure sets. DICOM imports. Standardized naming conventions are recommended. They’re not required, but if they are done properly, they’ll automatically be selected during patient import in AlignRT. Saves our therapists some confusion. Again, automate your patient imports when possible. It avoids any kind of complication. Daily QA, very simple for our clinics. We just check coincidence between the cameras each day. Monthly QA, a little bit more complicated. We check beam control and interlock functionality. Beam hold control, I meant to say. And gated beam output constancy, as well as the static localization accuracy. Annual QA adds on an end-to-end localization test and isocenter verification.

So real quickly, the unexpected. Commissioning is only as stringent as your tools allow. Phantom selection, this is where you might be caught by surprise. In our case, we used the tools that were available to us. Going back in hindsight, I recommend buying a phantom that’s more applicable to the treatment sites that you’re going to be using. This is one that we found later on. This is just an Amazon-available torso phantom. Reproduces a standard breast case. Allows you to look at the rigid registration transfer, something that mirrors the human anatomy a little bit better than a cube does. This also comes in different skin tones, which is great depending on your patient selection. Consider how you will model respiratory motion. And going back, IT infrastructure. Bring in your IT team, have them on call the day of acceptance, and your life will go much easier.

Closing remarks. Your task group is your guide, but design of commissioning checklist is up to physics. Expect pre-acceptance to take anywhere between weeks to months. Commissioning should take several days, upwards of a week, without much disruption of clinical flow. Include your whole clinical team, from your therapist to your dosimetrist, all the way through the commissioning workflow process. It builds competency and trust in the system. Build confidence one protocol at a time. Start with something simple, breath DIBH. Work your way to something more complex, kidney DIBH or something that you–. Don’t go live with 45 different kinds of treatment on day one. Your therapist will hate you. Remember that skin surface is not your target. You must quantify and correlate between your skin surface ROI and the deep seated anatomy of every patient. So with that said, thank you very much for your time, and I’d like to thank my team at Vanderbilt.