Preliminary report on longitudinal stability of MRI QA up to 2 years on 7 clinical 1.5 T MR-Linacs

Abstract

Purpose or Objective MR-guided RT leverages the excellent soft-tissue contrast of MRI for daily treatment adaptation, facilitating hypo fractionated and dose-escalated treatments. To deliver precision RT on MR-Linacs, geometric fidelity and long-term stability of MR imaging components is essential. These components differ from conventional MRI systems due to the need for radio-translucency and exposure to scattered ionizing radiation. We report preliminary data on temporal stability of periodic MR imaging QA on MR-Linacs, which could inform frequency and action levels. Materials and Methods Seven centers contributed longitudinal MRI QA data acquired over 3-24 months on their 1.5 T MR-Linac systems (Elekta AB, Stockholm). To ensure consistent data across institutes, only phantoms provided with the MR-Linac system were used and a prescribed scan protocol was followed. A large cylindrical phantom (37 cm diameter) was used to measure B0 and B1 homogeneity and B0 dependence on gantry position. In addition to a monthly axial measurement, coronal and sagittal scans were performed every three months. For B0 mapping, phase images were acquired in separate TRs with TE1/TE2=5.4/6.9 ms to minimize phase wrapping, while a double flip angle technique with α1/α2=60°/120° was used for B1 mapping. All tests were conducted at gantry angle 0° except gantry dependence for which phase images were acquired without re-shimming at TE=15.7 ms for 13 gantry angles from -180° to 180°. For a circular ROI of diameter 35 cm through the isocenter, we report mean, standard deviation and range from 1st to 99th percentile for each measurement. The vendor-provided PIQT phantom–a multi-purpose phantom with 20 cm diameter, similar to the ACR phantom–was used for weekly image scaling and SNR tests. Here we report on flood field uniformity and spatial linearity. A phantom consisting of 7 slabs with position markers was used to assess spatial accuracy of the MR images in a large field of view. We report change in mean and 99th percentile over time for the distance to known position for markers within a spherical DSV of 35 cm. Geometric inaccuracy was 0.48±0.24 mm with the 99th percentile being 1.21 mm. Results were very stable over time, with an average intra-site standard deviation of 0.03 mm. For one system, PIQT analysis found SNR decline of the transmit body coil, which was replaced after 10 months of operation. Mean values post-replacement differed significantly (p<0.0001). Conclusion The vendor-provided PIQT tests capture many parameters that can reveal performance decline of components, but measurements are limited to a small field of view. Considering the dosimetric importance of the patient outline, larger phantoms are required to characterize the geometric accuracy of the MR images. We found all QA tests highly stable over time unless hardware components were exchanged. Most tests were highly comparable between sites with the exception of B0 homogeneity as a function of gantry angle due to site-specific shimming.