Background: Optimal alignment is of utmost importance when treating pediatric patients with craniospinal irradiation (CSI), especially with regards to field junctions and multiple isocenters and techniques applying high dose gradients. Here, we investigated the setup errors and uncertainties for pediatric CSI using different setup verification protocols.
Methods: A total of 38 pediatric patients treated with CSI were identified for whom treatment records and setup images were available. The setup images were registered retrospectively to the reference image using an automated tool and matching on bony anatomy, subsequently, the impact of different correction protocols was simulated.
Results: For an action-level (AL)-protocol and a non-action level (NAL)-protocol, the translational residual setup error can be as large as 24 mm for an individual patient during a single fraction, and the rotational error as large as 6.1°. With daily IGRT, the maximum setup errors were reduced to 1 mm translational and 5.4° rotational versus 1 mm translational and 2.4° rotational for 3- and 6- degrees of freedom (DoF) couch shifts, respectively. With a daily 6-DoF IGRT protocol for a wide field junction irradiation technique, the residual positioning uncertainty was below 1 mm and 1° for translational and rotational directions, respectively. The largest rotational uncertainty was found for the patients' roll even though this was the least common type of rotational error, while the largest translational uncertainty was found in the patients' anterior-posterior-axis.
Conclusions: These results allow for informed margin calculation and robust optimization of treatments. Daily IGRT is the superior choice for setup of pediatric patients treated with CSI, although centers that do not have this option could use the results presented here to improve their margins and uncertainty estimates for a more accurate treatment alignment.
Keywords: Craniospinal irradiation; Image guidance; Positioning errors; Positioning uncertainties; Residual setup errors.