Characterization and longitudinal assessment of daily quality assurance for an MR‐guided radiotherapy (MRgRT) linac

Abstract Purpose To describe and characterize daily machine quality assurance (QA) for an MR‐guided radiotherapy (MRgRT) linac system, in addition to reporting a longitudinal assessment of the dosimetric and mechanical stability over a 7‐month period of clinical operation. Methods Quality assurance procedures were developed to evaluate MR imaging/radiation isocenter, imaging and patient handling system, and linear accelerator stability. A longitudinal assessment was characterized for safety interlocks, laser and imaging isocenter coincidence, imaging and radiation (RT) isocentricity, radiation dose rate and output, couch motion, and MLC positioning. A cylindrical water phantom and an MR‐compatible A1SL detector were utilized. MR and RT isocentricity and MLC positional accuracy was quantified through dose measured with a 0.40 cm2 x 0.83 cm2 field at each cardinal angle. The relationship between detector response to MR/RT isocentricity and MLC positioning was established through introducing known errors in phantom position. Results Correlation was found between detector response and introduced positional error (N = 27) with coefficients of determination of 0.9996 (IEC‐X), 0.9967 (IEC‐Y), 0.9968 (IEC‐Z) in each respective shift direction. The relationship between dose (DoseMR/RT+MLC) and the vector magnitude of MLC and MR/RT positional error (Errormag) was calculated to be a nonlinear response and resembled a quadratic function: DoseMR/RT+MLC[%] = −0.0253 Errormag [mm]2 − 0.0195 Errormag [mm]. For the temporal assessment (N = 7 months), safety interlocks were functional. Laser coincidence to MR was within ±2.0 mm (99.6%) and ±1.0 mm (86.8%) over the 7‐month assessment. IGRT position–reposition shifts were within ±2.0 mm (99.4%) and ±1.0 mm (92.4%). Output was within ±3% (99.4%). Mean MLC and MR/RT isocenter accuracy was 1.6 mm, averaged across cardinal angles for the 7‐month period. Conclusions The linac and IGRT accuracy of an MR‐guided radiotherapy system has been validated and monitored over seven months for daily QA. Longitudinal assessment demonstrated a drift in dose rate, but temporal assessment of output, MLC position, and isocentricity has been stable.

Another unique challenge of MRgRT is the ability to verify IGRT isocenter coincidence to radiation isocenter. Since conventional linear accelerators have a single on-board detector that is compatible with both imaging and radiation source, simple localization of a phantom through the two modalities (i.e., MV/kV) is used to verify coincidence, while MRgRT systems contain no such detector. Currently, no method of verification of isocentricity has been reported in the literature. Current solutions to evaluate MR/RT isocentricity for MRgRT systems have been practically carried out with film enclosed by a water phantom using a star shot irradiation technique, eliminating real-time information and impractical for a daily QA technique. 2 Daily QA guidance has been previously established for conventional linear accelerators and MR imaging systems. [3][4][5][6][7][8] The daily QA tests recommended in TG142, TG40, and MPPG 8a are designed to maintain safety, accurate patient localization, and dosimetric output by monitoring parameters which can impact treatment goals. 3,5,6 However, there is no existing literature or guidance for routine QA on the two integrated systems to ensure consistent and safe operation and accurate treatment delivery using MRgRT.
In this study, we have developed and implemented an efficient and sensitive QA procedure to characterize the MR imaging/radiation isocenter alignment, spatial fidelity of imaging and patient handling systems, and the performance of the linear accelerator on an MRgRT system for routine daily QA. A method for real-time characterization of the MR/RT isocentricity is established through exploiting the sensitivity of the penumbra position across a large detector relative to the field size. As such, our method utilizes an MR-compatible A1SL detector (Standard Imaging Inc., Middleton, WI, USA) with an active volume of 4.4 mm length and 4 mm diameter placed within a cylindrical water phantom with a 0.40 cm 2 x 0.83 cm 2 field size to optimize the spatial sensitivity of MR/RT isocentricity and MLC position.
This work is the first reporting of a daily QA procedure for an MRguided radiotherapy system. The sensitivity of our methods has been characterized through introducing known errors and/or through comparing to established QA procedures. Lastly, we describe the first reporting of the longitudinal assessment of the dosimetric and mechanical accuracy of a commercial MRgRT linac over a 7-month period of clinical operation.

2.A | ViewRay MRIdian linac
The MRIdian linac, previously described by Hill and Mittauer, consists of a gantry-mounted 6 MV linear accelerator and a 0.345 T MRI scanner. 2 The linac produces a 6-MV flattening filter free beam with a nominal dose rate of 600 MU/min. Beam collimation is achieved using the RayZR™ MLCs, consisting of a set of two banks of MLCs, stacked and double focused, and offset by one-half leaf width, eliminating the need for MLC tongue and groove design. 9 With a 90 cm SAD, the MLCs project to a maximum field size of 27.4 cm 2 x 24.07 cm 2 at isocenter with individual leaf width projections of 8.3 mm.

2.B | Overview of phantom and QA procedures
An MR-compatible A1SL detector (active volume of 4.4 mm length and 4 mm diameter) within a cylindrical water phantom (ViewRay Inc., Cleveland, OH, USA) shown in Fig. 1 was utilized for this study.
The phantom is filled with distilled water to enable MR imaging capabilities. The phantom includes scribed locations for laser alignment which are coincident with the centroid of the active volume of the ionization chamber. There are four additional chamber positions located at the periphery of the cylindrical phantom. The phantom is indexed to the table through two mounting brackets, with a cutout for the posterior-oriented torso receiver coils.
An overview of the daily QA procedures is listed in Table 1, categorized by dosimetry, mechanical and imaging, and safety tests, along with the applicable tolerance from TG142 based on SBRT/SRS specifications. A description of the QA method and the technique used to characterize the method is also listed in Table 1.

2.C | Safety functionality
Implemented safety checks include functionality of patient monitoring, radiation monitoring, beam interruption, and door interlocks.

2.D | MR spatial fidelity and phantom localization
The room lasers define a virtual isocenter located −155 cm in IEC-Y direction from the MRIdian MR/RT isocenter. To verify that the lasers are coincident to this virtual isocenter, the phantom is initially aligned to the in-room lasers using the external scribe marks, translated +155 cm in the IEC-Y direction, and then localized based on MR imaging. A balanced steady-state-free precession sequence (TrueFISP) MR scan is acquired in 65 s with a 1.5 mm 3 x 1.5 mm 3 x 1.5 mm 3 resolution over 45 cm 3 x 23 cm 3 x 26 cm 3 field of view for the daily setup MR scan and the simulation reference MR scan, that is, the primary dataset used for the treatment plan generation.
Maximum spatial distortion of the MRIdian TrueFISP sequence is <1 mm within 5 cm of isocenter. 10 Localization is then achieved through manual alignment of the chamber holder about the active volume of interest for the A1SL ionization chamber. Phantom shifts are recorded as the difference between laser and imaging isocenters for the IEC-X, IEC-Y, and IEC-Z dimensions.
Postonline couch shifts for initial phantom localization, an IGRT position-reposition test is performed in accordance with TG142.
Here, a known shift is introduced of −0.75 cm (IEC-X), −4.9 cm (IEC-Y), and +0.75 cm (IEC-Z), and the phantom is subsequently reimaged. Couch movement and geometric spatial fidelity of the MR imaging is assessed using the known physical landmarks within the phantom (Fig. 2). The image of the shifted phantom ideally places the edges of the peripheral chamber inserts in a known geometry.
Specifically, at the image volume origin, the IEC-X landmark intersects the edge of the sagittal plane in the axial view, and the IEC-Z landmark intersects the edge of the coronal plane in the same axial view (Fig. 2). The IEC-Y landmark is visually verified by identifying the beginning of its edge on the superior-adjacent axial slice (i.e.,

2.E | Dosimetry, MLC, and MR/RT isocentricity
Postphantom localization of the chamber active volume with isocenter, a five-field 3D conformal treatment plan ( Fig. 3) is delivered. The plan includes a 10.04 cm 2 x 9.96 cm 2 field delivered with the gantry at zero degrees (G0, IEC 1217) to measure the dosimetric output and a 0.40 cm 2 x 0.83 cm 2 field at each cardinal angle characterizes the spatial accuracy of MR/RT isocentricity and MLC positional accuracy at each cardinal gantry angle. This field size was selected to optimize spatial sensitivity with the A1SL active volume (described above

2.F | Characterization of methods
The sensitivity of our methods has been characterized through intro- The dosimetric output using the daily QA procedure was benchmarked by comparison to measurements performed with monthly QA using TG51 protocol in a water tank, in addition to an independent output verification through an Accredited Dosimetry Calibration Laboratory (ADCL) service with TLD irradiation. 12    The coefficient of determination (R 2 ) and root mean square error (RMSE) are displayed in Fig. 5 for the respective equations. Note that the measurements for shifts in the gun-target direction were  All safety QA procedures found in Table 1 passed functionality on a daily basis over the longitudinal assessment. F I G . 6. Comparison of dosimetric output stability for monthly QA (TG51 protocol in water tank), daily QA (cylindrical water phantom), and an independent output measurement (ADCL-reported TLD irradiation).
F I G . 7. Histogram of longitudinal assessment of laser positional accuracy to MR isocenter coincidence.

| DISCUSSION
The clinical efficacy of MR guidance has previously been shown by the MRgRT community. 13 The superior soft tissue visualization combined with real-time tracking capabilities has enabled greater confidence in the treatment delivery allowing for a reduction in the planning target margin compared to CT-based IGRT modalities. 13 However, the temporal assessment of a clinical MRgRT system or implementation of daily QA has yet to be reported. Current commercial daily QA equipment and guidance criteria are limited to CTbased IGRT modalities may not be applicable to MRgRT systems due to equipment incompatibility in a magnetic field and/or fundamental differences in the clinical utility of the technology. In this study, we describe a novel daily QA procedure that exploits the spatial sensitivity of the penumbra to evaluate the MLC positional accuracy and MR/RT isocentricity for an efficient, robust daily QA procedure. In addition, we report the first evaluation of the longitudinal assessment of a clinical MRgRT linac system in terms of IGRT spatial fidelity and linac integrity.
Our technique has allowed TG142 criteria to be applicable and quantified on a daily basis. Currently, no guidance has been established for best practices and/or tolerances for routine QA of clinical MRgRT systems. For our evaluation, we applied TG142 tolerances for CT-based IGRT systems where applicable. Daily QA techniques on CT-based modalities often employ on-board imaging systems. 1,14 However, the implementation of x-ray-based detectors may not be applicable to MRgRT with some commercial MRgRT systems not having an onboard x-ray detector (i.e., MRIdian), therefore an external array or ionization chamber is necessary. Our technique has employed one phantom in combination with a single ionization chamber. However, care must be taken with daily handling of an ionization chamber and triaxial cable.
Implementation of the daily output was benchmarked by comparing two independent procedures: TG51 protocol in water tank,  installation (Fig. 9). The dose rate was intentionally increased in July 2018 through changing the PRF from 161 Hz to 183 Hz. The increase of PRF was noted to have a small increase in the linac dark current, which continues to be monitored on a monthly basis.
The MR/RT isocentricity and MLC positional accuracy as measured with 0.40 cm 2 x 0.83 cm 2 field at each cardinal angle was benchmarked for sensitivity through introducing known errors across IEC-X, IEC-Y, IEC-Z for each cardinal angle (Fig. 5). As such, implementation of this small field allowed for characterization of not only shifts due to MR/RT offset but also MLC positional error as a function of gantry angle. The overall sensitivity was found to be 5.6% (1 mm) and 24.7% (2 mm) for IEC-X [ Fig. 5(a)  A limitation of the sensitivity of MR/lasers and IGRT positioning and repositioning evaluation is the spatial resolution of the imaging sequence. We used the highest available resolution at 1.5 mm 3 x 1.5 mm 3 x 1.5 mm 3 . Due to volume averaging across voxels and gantry dependency of the MRI, some resulting deviations were greater than 1 mm, however, 99.4% of measurements remained less than 2 mm.
This approach of setting the gantry angle during MR imaging to a specified location to minimize MR/RT isocentricity has been implemented clinically at our institution for the initial 3D volumetric scan for MRgRT patient setup and simulation images. 15 For IGRT position-reposition accuracy, a greater drift was noted at 10/2018 to 01/ 2019 (Fig. 8), likely attributed to interuser dependence of phantom localization, as timepoint corresponds to a rotation of operator.
Although not performed for this study, mitigation of MRI gantry dependence can be performed through 3D volumetric imaging at the optimal gantry angle in which the centroid of MR isocenter is equal to the centroid of RT isocenter. Such an approach has been implemented clinically for the initial 3D volumetric scan for MRgRT setup and simulation images at our institution.

Additional unique considerations for MRgRT systems include
fidelity of the MR imaging system, that is, spatial integrity and overall functionality of the MRI scanner. Verification of the absence of any ferromagnetic objects being lodged in the MRI scanner and communication between the MRI scanner and radiotherapy user interface are necessary components of the daily QA. Through our implemented technique, the daily QA serves as an end-to-end procedure in the clinical treatment workflow, therefore enabling evaluation that the TPDS and MRI scanner are communicating in the clinical state. An additional consideration with regard to the overall health of the MRI scanner is the integrity of the MRI receiver coils.
At our institution a weekly procedure to verify the signal to noise (SNR) and percentage integral uniformity for coil integrity is performed. One could easily incorporate coil robustness by measuring the SNR over the uniform water areas of the phantom for the daily QA procedure.
A limitation of our technique is that the real-time tracking and online adaptive components are not incorporated. Both mechanisms are likely low failure frequency as has been demonstrated over institutional practice and QA. Nonetheless, QA of these components should be established in a routine QA program. One practical method to verify treatment planning system integrity for online adaptive planning is incorporating a checksum to identify unintentional modifications to system configurations and database contents. 4,16 Detailed account of an end-to-end validation of online adaptive radiotherapy has been previously described by Mittauer et al. for an MRgRT program. 17 Quality assurance of the gating capabilities of the MRIdian can be monitored with an end-to-end procedure with a motion phantom as described by Lamb et al. 18 Although not part of TG142 criteria for daily QA, a limitation of our technique is that energy is not verified on a daily basis. One method to incorporate an energy check into the existing phantom design is to add a second ionization chamber at the IEC-X landmark, that is, chamber holder. Here, the ratio of the distal chamber to the central chamber for the 10.04 cm 2 x 9.96 cm 2 field at G0 could be evaluated as a surrogate for energy constancy.