Brain stereotactic radiosurgery using MR-guided online adaptive planning for daily setup variation: An end-to-end...

Online magnetic resonance (MR)-guided radiotherapy is expected to benefit brain stereotactic radiosurgery (SRS) due to superior soft tissue contrast and capability of daily adaptive planning. The purpose of this study was to investigate daily adaptive plan quality with setup variations and to perform an end-to-end test for brain SRS with multiple metastases treated with a 1.5-Tesla MR-Linac (MRL). The RTsafe PseudoPatient Prime brain phantom was used with a delineation insert that includes two predefined structures mimicking gadolinium contrast-enhanced brain lesions. Daily adaptive plans were generated using six preset and six random setup variations. Two adaptive plans per daily MR image were generated using the adapt-to-position (ATP) and adapt-to-shape (ATS) workflows. An adaptive patient plan was generated on a diagnostic MR image with simulated translational and rotational daily setup variation and was compared with the reference plan. All adaptive plans were compared with the reference plan using the target coverage, Paddick conformity index, gradient index (GI), Brain V12 or V20, optimization time and total monitor units. Target doses were measured as an end-to-end test with two ionization chambers inserted into the phantom. With preset translational variations, V12 from the ATS plan was 17% lower than that of the ATP plan. With a larger daily setup variation, GI and V12 of the ATS plan were 10% and 16% lower than those of the ATP plan, respectively. Compared to the ATP plans, the plan quality index of the ATS plans was more consistent with the reference plan, and within 5% in both phantom and patient plans. The differences between the measured and planned target doses were within 1% for both treatment workflows. Treating brain SRS using an MRL is feasible and could achieve satisfactory dosimetric goals. Setup uncertainties could be accounted for using online plan adaptation. The ATS workflow achieved better dosimetric results than the ATP workflow at the cost of longer optimization time.