Longitudinal Stability of MRI QA up to Two Years on Eight Clinical 1.5 T MR-Linacs

MR-guided radiotherapy on hybrid MR-Linacs exploits the excellent soft-tissue contrast of MRI to deliver daily adaptive precision radiotherapy. Geometric fidelity and long-term stability of MRI components are essential, but their longitudinal performance under daily exposure to scattered ionizing radiation is unknown. We report on longitudinal stability of periodic MRI QA on eight clinical 1.5T MR-Linac systems. We provided measurement instructions for periodic MRI QA and received data from seven different centers within the Elekta MR-Linac consortium, which contributed data over acquisition periods ranging from 3-24 months. We tested B-0 and B-1 homogeneity using a 37 cm diameter cylindrical phantom, which was measured monthly in axial orientation, supplemented by quarterly sagittal and coronal acquisitions. We report average, standard deviation and peak-to-peak variation (99th-1st percentile) within a region of interest (ROI) of 35 cm diameter. Dependence of B-0 on the gantry angle and gradient non-linearity were tested quarterly. We analyzed the longitudinal stability of selected metrics of the vendor-provided periodic image quality tests. We found high temporal stability of B-0 and B-1 measurements and good agreement between different MR-Linac systems. For all measurements, the standard deviation of B-0 within the analyzed ROI was below 0.66/0.33/0.33 ppm for axial/sagittal/coronal orientation. The average standard deviation of the ratio between actual and nominal flip angle was 0.022/0.100/0.088 for axial/sagittal/coronal orientation. Systems exhibited distinctively different gantry angle dependencies of B-0, with sensitivities of B-0 to the gantry angle differing by factors of up to two between systems. Gradient non-linearity analysis yielded average radii of 172 and 242 mm for which 98% of the phantom markers had deviations below 1 and 2 mm, respectively. All analyzed periodic image quality tests were passed, but major events including a body coil replacement and ramp down were apparent in the time series. Overall we found very similar performance of the tested systems and our results could inform the implementation of MR imaging QA for MR-Linacs. While we found differences of the gantry angle dependence of B-0 between systems, the high temporal stability found for all tests is a foundation for stereotactic radiotherapy and multi-center clinical trials involving quantitative MRI.