Purpose: This work outlines a robust solution to MR-Linac daily QA, including mechanical-imaging-radiation isocenter coincidence verification using Cherenkov imaging.
Methods: A fully enclosed and sealed acrylic cylindrical phantom was designed to be mountable to the manufacturer-provided jig. A custom-machined plastic conical structure was fixed inside the phantom and held in place with 3D-printed spacers. The phantom was filled with water, allowing for high edge contrast of the plastic cone on MR images. Both a star shot plan and a four-angle sheet beam plan were developed and delivered to the phantom; the former allowed for radiation isocenter localization in the x-z plane (A/P and L/R directions) relative to physical landmarks on the phantom, and the latter allowed for the longitudinal position of the sheet beam to be encoded as a ring of Cherenkov radiation emitted from the phantom, allowing for isocenter localization on the y-axis (S/I directions). A custom software application was developed in MATLAB to perform near-real-time analysis of the data.
Results: Calibration procedures showed that linearity between longitudinal position and optical ring diameter is high (R² > 0.99), and that RMSE is low (0.184 mm). The star shot analysis showed a minimum circle radius of 0.34 mm. The final isocenter coincidence measurements in the lateral, longitudinal, and vertical directions were -0.61 mm, 0.55 mm, and -0.14 mm respectively, and the resulting 3D coincidence was 0.83 mm, below the 2 mm tolerance defined in TG-142. Longitudinal analysis showed an average coincidence of 1.5 mm ± 0.4 mm over 8 weeks of daily use.
Conclusion: This novel work highlights an efficient method for acquisition and near-real-time analysis of MR-Linac isocenter coincidence data and represents a direct measurement of the 3D isocentricity. This phantom and custom analysis application combined with longitudinal analysis makes this solution easily transferrable to clinical use.
Funding Support, Disclosures, and Conflict of Interest: This work has been funded by NIH grant R01EB023909. Brian Pogue and Petr Bruza are affiliated with DoseOptics LLC, which provided hardware support for this study. Daniel Alexander is a research consultant for DoseOptics LLC outside the context of this work.