Planning Target Volume Implications of Residual Setup Uncertainty and Intrafraction Motion During MRI Guided Brain Radiotherapy.

PURPOSE/OBJECTIVE(S) Magnetic resonance image guided brain radiotherapy (MRIgRT) is a promising advance to maximize the therapeutic ratio. An unexplored uncertainty in this approach is the residual setup uncertainty after translation-only MR based patient setup. This work quantified this uncertainty, including intrafraction motion, and determined the impact on planning target volume (PTV) margins. MATERIALS/METHODS Sixty-six patients treated on a 1.5T-based MR-Linac were included. Of these, 49 and 17 patients received conventional (15-30 fractions for glioblastoma) and hypofractionated (5 fractions for post-operative brain metastases) treatments, respectively; 1,329 total patient fractions were analyzed. At each fraction, patients were setup using online MR guidance by translation-only (T) fusion of the T1 volumetric MR to the planning image set to determine the treatment isocenter shift. These fusions were independently validated offline with translational and rotational (T+R) fusion of bony anatomy. The difference between the T and T+R co-registrations was applied to the original clinical target volume (CTV) to yield the CTV at each fraction (CTVtreat). The set of all CTV and CTVtreat volumes was incorporated in a population-based model parameterized by the relative thresholds (CTVmin, Fxmin, Ptmin). This model determined the minimum planning target volume (PTV) margin such that the PTV encompassed CTVmin of the CTVtreat volume in ≥ Fxmin of fractions in ≥ Ptmin of patients. In this work, we determined the PTV margin that for two sets of thresholds: A) (CTVmin, Fxmin, Ptmin) = (98%, 95%, 90%) and B) (CTVmin, Fxmin, Ptmin) = (95%, 95%, 90%). In a subset of 412 fractions, intrafraction motion was determined as the spatial registration difference between a post-treatment acquired MRI and the pre-treatment MRI. RESULTS Residual setup uncertainty (difference between T+R and T registrations) and intrafraction motion results are summarized in Table 1. Across all patient fractions, the vector magnitude of the setup uncertainty and intrafraction motion was 1.3 ± 0.8 mm (mean ± SD) and 0.6 ± 0.5 mm, respectively. To accommodate the setup uncertainty, PTV margins of 2.5 and 1.8 mm were required to meet criteria sets A and B, respectively. CONCLUSION We have quantified residual setup and motion uncertainties for a large series of treated MR-Linac brain tumor patients, generated a model for a population-based PTV, and recommend a minimum PTV of 2-3 mm for this indication. This work is a crucial step towards developing an adaptive brain treatment based on dynamic tumor changes and tumor response.