Inter-fraction Cardiac Substructure Displacement Assessed via Magnetic Resonance-guided Radiation Therapy

Purpose/Objective(s): Growing evidence suggests that radiation therapy (RT) doses to the heart and cardiac substructures are strongly linked to cardiac toxicities. However, it is challenging to define substructures on conventional non-contrast treatment planning CTs due to limited soft tissue contrast. This work leverages unique longitudinal MR-guided RT data to evaluate inter-fraction displacements of cardiac substructures to facilitate safety margin design. Materials/Methods: Nine patients with non-breast intrathoracic malignancies treated with MR-guided RT using a 0.35T MR-linac were identified for this study. Of these, 8 patients were treated at breath-hold (6 end-inhalation and 2 end-exhalation using a 17-25 second TrueFISP scan with 1.5×1.5×3 mm3 resolution) and 1 under free breathing conditions (∼3 minute TrueFISP scan with 1.5×1.5×1.5 mm3). Daily MRIs (n = 35 total for the cohort) obtained in the same breath-hold conditions were rigidly registered to the planning reference MR-simulation (MR-SIM) datasets using bony alignment. A refined translational registration surrounding the planning target volume was applied using normalized mutual information as the similarity metric. A hybrid diagnostic MR/CT atlas was used to automatically propagate contours for the heart and 12 cardiac substructures (chambers, coronaries, and great vessels) to all serial MRIs. Propagated contours were modified by two radiation oncologists as needed. Inter-fraction differences were quantified via centroid analysis and dominant axes of motion were identified for each substructure. Results: Over all fractions, the heart inter-fraction shift from the reference MR-SIM was70%, 78%, and 64% of the fractions in the left-right (LR), anterior-posterior (AP), and superior-inferior (SI) directions, respectively. Left and right ventricles had a tendency for larger inter-fraction displacements in the LR and SI directions (median shifts of ∼2.7-2.8 mm, ∼51-54% of fractions \u3c 3 mm) as compared to ∼1.8 mm in the AP direction. Pulmonary veins showed substantial differences (median ∼5 mm) in the LR as compared to the other dimensions (∼2 mm). The ascending aorta shifted similarly in all axes (median shifts 1.8-2 mm). The left anterior descending artery had the highest AP shift (median ∼6 mm) of all substructures evaluated. For two unique patient fractions, SI deviations were \u3e1 cm for most substructures due to lack of compliance with breath-hold conditions. Conclusion: Individual cardiac substructure displacement demonstrated variability in magnitude and dominant axis, suggesting that anisotropic substructure-specific planning organ at risk margins may be warranted. With confirmation in a larger cohort stratified by respiratory state, precise margins can be derived including systematic and random uncertainties for more robust cardiac substructure sparing