Risk and Intervention Assessment for Rotational and Translational Setup Errors on Spinal Cord Dose for Stereotactic Body Radiotherapy (SBRT) of Paraspinal Tumors: An Automated Planning Study

Abstract

Translational image-guided setup corrections for SBRT of paraspinal tumors ensure geometrical accuracy at isocenter but do not address rotational errors. The objectives of this work are to assess the impact of combined residual rotational and translational errors on spinal cord dose and determine factors that predict cord dose sensitivity to these errors. Nine-field IMRT plans (18 Gy/1 fraction) were generated retrospectively for 16 spinal metastases patients with the objective of maximizing target coverage while keeping the dose to 0.1cc of the cord to 10Gy (Dcord). The maximum dose to 0.1cc of the volume consisting of the cord + 2mm margin (DPRV2) was recorded. To assess the impact of rotational error on Dcord, the treatment planning system was used in batch mode to recalculate each plan after introducing simulated combinations of rotational errors around the isocenter (0-12° around X, Y and Z-axis). These simulations were repeated after adding 1 mm residual translational errors along both X and Y-axis such that the cord was brought closer to the high dose region. Rotations resulting in the actual cord dose equal to DPRV2 were deemed significant. Rotation combinations were reduced to a single equivalent rotation (θeq) around an arbitrary axis using eigenvector formalism, and the minimum θeq for which Dcord = DPRV2 was compiled for each patient. Two predictive factors for Dcord sensitivity to rotational errors were evaluated using regression analysis. 3422 rotational and translational error combinations were evaluated. Without any translational shift, rotations where θeq = 2.6°-10.2° were required for Dcord to reach DPRV2 (12.74-14.98 Gy). In combination with a 1 mm residual translational shift, θeq decreased to 1.4°-7.7°. The strongest predictor for cord dose sensitivity to rotational errors was the maximum distance between the isocenter and the center of the cord cross-section on the most inferior or superior CT slices with 95% isodose (diso-cord). A linear model based on the first 8 patient plans using diso-cord and the dose gradient at the target-cord interface was able to predict θeq within ± 0.9° (standard deviation) for the next 8 patients. Simulations of combined setup errors in paraspinal SBRT plans showed limited cord dose sensitivity to rotational errors and demonstrated the importance of minimizing residual translational errors. A linear model predicting cord dose sensitivity to rotational errors was designed and will be used to evaluate intervention strategies such as the six-degree of freedom couch.