Implementation of free breathing respiratory amplitude‐gated treatments

Abstract Purpose The purpose of this study was to provide guidance in developing and implementing a process for the accurate delivery of free breathing respiratory amplitude‐gated treatments. Methods A phase‐based 4DCT scan is acquired at time of simulation and motion is evaluated to determine the exhale phases that minimize respiratory motion to an acceptable level. A phase subset average CT is then generated for treatment planning and a tracking structure is contoured to indicate the location of the target or a suitable surrogate over the planning phases. Prior to treatment delivery, a 4DCBCT is acquired and a phase subset average is created to coincide with the planning phases for an initial match to the planning CT. Fluoroscopic imaging is then used to set amplitude gate thresholds corresponding to when the target or surrogate is in the tracking structure. The final imaging prior to treatment is an amplitude‐gated CBCT to verify both the amplitude gate thresholds and patient positioning. An amplitude‐gated treatment is then delivered. This technique was commissioned using an in‐house lung motion phantom and film measurements of a simple two‐field 3D plan. Results The accuracy of 4DCBCT motion and target position measurements were validated relative to 4DCT imaging. End to end testing showed strong agreement between planned and film measured dose distributions. Robustness to interuser variability and changes in respiratory motion were demonstrated through film measurements. Conclusions The developed workflow utilizes 4DCBCT, respiratory‐correlated fluoroscopy, and gated CBCT imaging in an efficient and sequential process to ensure the accurate delivery of free breathing respiratory‐gated treatments.

visibility and is most appropriate for large lung lesions or cases where there is a high-density surrogate, such as fiducials. 4D cone beam computed tomography (4DCBCT) correlates CBCT projections with the measured respiratory trace to reconstruct volumetric images at different parts of the breathing cycle. 10,11 This reduces respiratory motion artifacts and offers improved verification of the ITV compared to 3D CBCT, which has been shown to underestimate the ITV in certain cases. 12 Gated CBCT is another valuable pretreatment imaging technique that acquires CBCT projections in a predefined gating window. This is useful to mitigate motion artifacts and verify the gating thresholds prior to the delivery of gated treatments. 13 Free breathing gating is a respiratory motion management strategy that allows patients to breathe normally during treatment while radiation is delivered during a predefined portion of the respiratory cycle. Phase-and amplitude-based approaches have both been used to gate the beam to minimize target motion while radiation is delivered. 14,15 It has previously been suggested that such techniques require respiratory-gated imaging to ensure accurate treatment delivery. 9 As pretreatment imaging techniques advance there is the opportunity to improve the accuracy of free breathing gated treatment deliveries and account for changes in respiratory motion from time of simulation to treatment.
The implementation of free breathing gated treatment programs requires a dedicated commissioning process. However, there are few recommendations in the literature regarding workflow and commissioning best practices. TG-142 offers several QA recommendations for gated systems including beam output and energy constancy, gate temporal accuracy, and surrogate calibration. 16 Methodology to perform such tests has previously been reported in the literature. 17 However, as gated treatment deliveries evolve with advances in pretreatment imaging, the tests performed during commissioning must follow suit.
The purpose of this work is to provide a practical guide to implementing free breathing amplitude-gated treatments using 4DCT, 4DCBCT, gated CBCT, and respiratory-correlated fluoroscopy. The workflow developed at our institution is described in detail, along with the steps taken to commission the technique. The pretreatment imaging process allows for the accurate delivery of free breathing amplitude-gated treatments and is robust to changes in breathing and respiratory motion from time of simulation to treatment.

2.A.1 | Phantom
The workflow for the delivery of free breathing respiratory-gated treatments was developed and commissioned using an in-house lung motion phantom, 18

2.A.2 | Motivation for amplitude gating
A key decision to make when developing a free breathing respiratory-gated treatment program is whether phase-or amplitude-based gating will be used. This decision is largely dependent on the capabilities of the available equipment and each technique's ability to handle irregular respiratory traces. The Siemens Definition AS20 CT (Siemens Healthineers, Erlangen, Germany) scanners at our institution are equipped with the Varian RPM system (Varian Medical Systems, Palo Alto, CA) and generate phase-based 4DCT scans. This evaluation motivated the use of amplitude-gated treatment deliveries with the predictive filter disabled to ensure that radiation is consistently delivered at patient exhale in the case of irregular breathing.

2.B | Clinical workflow
The clinical workflow implemented at our institution utilizes a marker block placed on the patient's abdomen as a surrogate of respiratory motion. A 4DCT is acquired at time of simulation and a phasebased reconstruction allows for analysis of target motion throughout the breathing cycle. A subset of phases over exhale where target motion is minimized to an acceptable amount is chosen for planning and treatment. Pretreatment imaging defines and verifies amplitude gate thresholds to correspond to the planning phases and an amplitude-gated treatment is delivered. This workflow has been clinically applied to a variety of disease sites at our institution, including lung, liver, cardiac, and abdominal cases. It is described below in detail and illustrated using a representative clinical lung case that was treated to 60 Gy in eight fractions.

2.B.1 | Simulation
A Siemens Definition AS20 CT scanner (Siemens Healthineers, Erlangen, Germany) is used to acquire a 4DCT scan at simulation.
The respiratory trace is recorded using the Varian RPM system It is important to understand the conventions that the 4D modalities use to label phases. The CT scanner used in this work defines 0% to be maximum inhale and reconstructs phases in ten 10% bins that are labeled as the minimum phase in that bin. For example, the 0% phase bin dataset includes phases 0%-9%, the 10% phase bin dataset ranges from phases 10% to 19%, the 90% phase bin dataset contains phases 90%-99%, and so forth. This is illustrated in Fig. 3a.
An example of 4DCT motion analysis is shown in Fig. 4 for the clinical lung case. There is a total of 2.5 mm, 6.5 mm, and 12.5 mm of motion in the right-left, anterior-posterior, and superior-inferior directions, respectively. Since superior-inferior motion dominates, we first look at the target centroid positions in that direction to determine which phases to treat over. Over the phase interval 20-60%, the minimum target centroid position is 9.2 mm (at phase 20%)

2.B.3 | Pretreatment imaging
A Varian TrueBeam v2.7 (Varian Medical Systems, Palo Alto, CA) linear accelerator is used for treatment and the TrueBeam reflector block is placed on the patient's abdomen to record a respiratory trace. The system first learns the patient's breathing trace over four respiratory cycles to set a baseline. Any couch movement greater than 2 mm causes the system to relearn and re-baseline the breathing trace.
The pretreatment imaging workflow is designed to determine and verify the amplitude gate corresponding to when the target or surrogate is inside the tracking structure generated during planning. The first imaging step is an orthogonal kV x-ray pair to provide an initial bony alignment. Next, a 4DCBCT scan is acquired. Using default settings, the 4DCBCT takes 2 minutes to acquire and 90 seconds to reconstruct. The 4DCBCT scan acquires projections throughout the respiratory cycle and, by default, reconstructs ten 10% phase bins. Maximum inhale is defined as 0%, and each phase dataset is labeled as the middle phase in that bin. For example, the 0% phase bin dataset includes phases 95%-4%, the 10% phase bin dataset ranges from phases 5% to 14%, etc. This is illustrated in Fig. 3b.
There are two 4D reconstruction algorithms available with the TrueBeam software: an "Advanced 4D" algorithm using a McKinnon-Bates reconstruction 19 and a "Basic 4D" algorithm. Briefly, the advanced algorithm uses all projections to reconstruct one 3D volume which is then forward projected to create difference projections that are compared to the acquired projections. The presence of motion causes dissimilarities between the difference and acquired projections, which are then added to the prior image for each phase bin dataset. The "Basic 4D" algorithm simply bins the acquired projection into each phase bin and reconstructs each phase bin image separately. It is common for 4DCBCT data to be undersampled in order to achieve a reasonable scan time, which comes at the cost of streaking and view aliasing artifacts. The "Advanced 4D" algorithm performs better than the "Basic 4D" algorithm on undersampled data. 19 Unless otherwise specified, all 4DCBCT scans acquired in this work used the "Advanced 4D" algorithm.
After the acquisition of the 4DCBCT, a subset average CBCT is reconstructed over the planning phases to match to the planning CT.
Due to the different method of phase binning between the 4DCT and 4DCBCT imaging systems, if the planning phase bins are x%-y %, the 4DCBCT average is reconstructed over phase bins (x + 10)%-y%. In the clinical lung example, the planning CT is over phase bins 20%-60% (which is individual phases 20%-69%) and the average CBCT for matching is reconstructed over phase bins 30%-60% (which is individual phases 25%-65%). If the 20% phase bin was included in the 4DCBCT phase subset average this would include phases 15-19%, which are outside of the planning phases.
A soft tissue match 20 is performed between the planning CT and the 4DCBCT subset average over the planning phases. Fig. 7 displays 4DCBCT image slices for the clinical lung case. In Fig. 7a, it can be seen that the target is entirely in the ITV on the 4DCBCT subset average, as expected. After this match, the 4DCBCT movie loop workspace is used to step through and assess target motion over all ten trace at which it was acquired. This is visualized in Fig. 8 for the clinical lung case. Using the tracking structure, which in this case is the ITV, the amplitude gate thresholds are adjusted such that the target is inside the tracking structure during the gate (contour turns yellow as in Fig. 8b) and the target is outside of the tracking structure outside of the gate thresholds (contour is green as in Fig. 8c). Evaluating when the target is in the tracking structure is a manual process and is determined jointly by the radiation oncologist, physicist, and radiation therapist. The same principle applies if a surrogate (e.g. fiducials, diaphragm) is used for motion assessment instead of the target.
The final imaging prior to treatment is an amplitude-gated CBCT.

2.C | Commissioning measurements
The lung motion phantom described in Section 2.A. was used for commissioning this technique. Two periodic breathing traces, shown in Fig. 9a, were programmed for commissioning. One trace was con-

2.C.1 | 4DCBCT commissioning
Evaluating the functionality of 4DCBCT was a key part of our commissioning process. Since this workflow requires using multiple systems to acquire 4D scans, it is crucial to understand and verify how each is binning phases. As discussed in Sections 2.B.1 and 2.B.3 and illustrated in Fig. 3, the 4DCT and 4DCBCT systems used in this work both divide the respiratory cycle into ten phase bins, but bin phases in different ways. To verify proper binning of the phases, the 4DCT and 4DCBCT images acquired using the lung phantom with both the 5 s/1 cm and 3 s/2 cm traces were evaluated to determine F I G . 6. The pretreatment imaging workflow used for alignment and to set the amplitude gate thresholds for treatment. Varian indicates that the "Advanced 4D" algorithm does not correctly represent the motion of radio-opaque fiducial markers, and therefore recommends using the "Basic 4D" algorithm for such cases. This was verified by placing two fiducials 2 cm apart in a cork lung insert and using the 5 s/1 cm trace to acquire 4DCBCT scans reconstructed with both available algorithms.

2.C.2 | End to end testing
For end to end testing, the entire workflow from simulation to treatment delivery was followed for both the 5 s/1 cm and 3 s/2 cm periodic breathing traces. Motion was evaluated and a phase subset average planning CT over phase bins 20-70% was reconstructed for each case, minimizing motion to 0.4 cm and 0.8 cm over the treated phases for the 5 s/1 cm and 3 s/2 cm cases, respectively. An ITV was contoured over the 20-70% phase subset average, with a 5 mm margin added to form the PTV. Simple AP/PA plans using static MLC-shaped 6 MV beams were developed to deliver 2 Gy to the PTV. These 3D plans were used to mitigate potentially confounding interplay effects present in VMAT plans. The pretreatment imaging workflow described in Section 2.B.3 was followed using the ITV as the tracking structure to determine amplitude gate thresholds.
Film measurements were performed with Gafchromic EBT3 film and analyzed with FilmQA Pro software (Ashland, Bridgewater, NJ, USA) following a previously published single scan, triple channel dosimetry protocol. 21 Film was placed inside of the cork lung insert in a coronal plane through the center of the solid water lesion.
Gamma analysis was used to compare film measurements to the planned dose distributions. 22 3 | RESULTS     Fig. 12a and Fig. 12b for the 5 s/1 cm and 3 s/ 2 cm cases, respectively. Film and planned dose distributions were also compared using gamma analysis with a pass rate criteria of 2%/ 2 mm and minimum threshold of 10%. The gamma pass rates were 98.6% and 98.5% for the 5 s/1 cm and 3 s/2 cm plans, respectively.

3.B.2 | Interuser variability
Interuser variability of the workflow was evaluated by having three independent users perform the pretreatment imaging and treatment delivery process with the 5 s/1 cm trace. While there is potential subjectivity in setting amplitude gate thresholds from the fluoroscopy images, the upper gate thresholds set by all three users were within 0.9 mm (4.1 mm, 4.5 mm, and 5.0 mm). Fig. 13 shows superior-inferior profiles extracted from each user's film measurement. When 2D film measurements from each user were compared to each other all combinations agreed with 2%/2 mm gamma pass rates of 99%.

3.B.3 | Robustness to irregular breathing
Film measurements were acquired using the irregular traces shown in Fig. 9b to determine the robustness of the technique to non-periodic breathing and changes in breathing between 4DCT acquisition and treatment. Fig. 14 shows superior-inferior profiles extracted from film measurements acquired with the nominal 5 s/1 cm and irregular breathing traces. The 2%/2 mm gamma pass rate when comparing the nominal and irregular trace measurements was 98%.
This indicates that this technique is robust to variable breathing patterns.   26 Therefore, our conservative policy in implementing this technique is to require reimaging when the breathing trace is relearned during treatment delivery to verify the relationship between the respiratory surrogate (TrueBeam reflector block) and internal anatomy.

| DISCUSSION
The TrueBeam system automatically relearns the breathing trace after any couch movement greater than 2 mm, whether it be due to shifts applied from imaging or couch centering. If the patient's breathing pattern changes or is irregular during the relearning period, there is the potential to re-baseline the respiratory trace at a different level compared to previous imaging. Therefore, our implementation of this technique requires that the last form of imaging prior to starting or resuming treatment is a gated CBCT where shifts are not required. Isocenter is placed to ensure that a gated CBCT can be acquired without centering the couch.
It has been demonstrated that patient coaching during treatment can improve breathing regularity and treatment reproducibility. 27,28 There are several audiovisual coaching options available within the TrueBeam system. Visual coaching can be provided in the form of a slider bar or movement along a path to indicate breathing amplitude.
Audio instruction is also available to instruct patients when to inhale and exhale. Both audio and visual feedback settings can either be set manually or tied to the patient's breathing pattern during the learning period. While these features have not been extensively explored at our institution, this could be used to help patients that do not naturally have a regular respiratory pattern.
While this work uses a TrueBeam linac, the developed workflow can be applied to other systems/vendors with similar capabilities, namely 4DCBCT, respiratory-correlated fluoroscopy, and gated CBCT. A particularly important aspect to consider when using other systems is the convention by which the 4DCT and 4DCBCT phases are binned to ensure that the phase subset average CBCT appropriately corresponds to the planning phase subset average CT. The commissioning process described here could be undertaken with other motion phantoms. The phantom used in this work was limited to superior-inferior motion, but it could be advantageous to use a phantom capable of 3D target motion to better represent anatomical motion.

| CONCLUSION
The free breathing respiratory-gated treatment program described in this work uses multiple imaging modalities in a logical and sequential pretreatment imaging workflow. 4DCBCT and fluoroscopy imaging allow for daily evaluation of motion and determination of amplitude gate thresholds. The commissioning process established the accurate end to end delivery of these treatments for multiple respiratory patterns and across different users.

AUTHOR CONTRI BUTIONS
Susannah Hickling, Andrew Veres and Michael Grams conceptualized the work and performed commissioning measurements. All authors contributed to data analysis and interpretation. Susannah Hickling prepared the manuscript. All authors critically revised the work and have given approval for the submission of this manuscript.

CONFLI CT OF INTEREST
No conflicts of interest.

D A T A A V A I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.