ESTRO 38 Abstract book

S530 ESTRO 38

robustness of our extreme hypofractionated optimized proton plans, with negligible effect on target coverage and bladder dose. Doses to the anorectum and anal canal were most sensitive to posterior and inferior prostate motion. Our results suggest that ERB has no additive value on plan robustness. Further investigation with a larger patient cohort is needed to confirm our results. PO-0973 Residual intra-fraction error in non- immobilized patients treated with tracked robotic spinal SBRT E. Rossi 1 , C. Fiorino 1 , A. Fodor 2 , G.M. Cattaneo 1 , C. Deantoni 2 , A. Sbalchiero 2 , A. Tavilla 2 , N.G. Di Muzio 2 , R. Calandrino 1 , S. Broggi 1 1 San Raffaele Scientific Institute, Medical physics, Milan, Italy ; 2 San Raffaele Scientific Institute, Radiotherapy, Milan, Italy Purpose or Objective To assess residual intra-fraction patient motion in the delivery of spinal radiosurgery with CyberKnife (CK, Accuray, Inc.) using spine tracking in non-immobilized patients. Moreover, the intra-fraction motion with respect to the first alignment (i.e. tracking not available) is quantified and its relationship with time is assessed. Material and Methods The Xsight Spine tracking technique with CK enables to track skeletal structures near the spine, avoiding the implantation of fiducial markers and, possibly, immobilization systems, allowing comfortable set-ups for, often, painful patients. X-ray images are taken to correctly align the patient on the treatment couch, then images are regularly acquired during treatment to correct for patient motion. Residual error during tracking deals with the shifts between consecutive images and may be safely quantified considering the registered shifts between them. We collected delivery data from 20 patients previously treated with tracking for spinal lesions without immobilization (51 fractions, 1615 images): 16-30 Gy (median 24Gy) in 1-5 fractions (10/20 with single fraction, 3/20 with multiple lesions) were delivered. The time between images for tracking varies between 30’’ and 2’ (median 1’ ± 30’’). Residual error is quantified for each image as the difference between measured translational and rotational corrections and the previous values. The error distribution for each fraction, patient and the whole population is assessed, including a possible relationship with time. Similarly, for intra-fraction motion in absence of tracking, we consider the shifts from the alignment coordinates (at time 0). Results Residual intra-fraction error after tracking is limited: translational shifts >0.5mm are detected in only 2.5%, 6.8% and 4.8% of 1615 acquired images for cranio-caudal, left-right and anterior-posterior translations respectively; shifts >1mm in 0.8%, 2%, 1.4%; shifts >1.5mm in 0.2%, 0.5%, 0.3%; shifts >2mm in 0.2%, 0.2% and 0.1%. As for roll, 0.7% of shifts are >1°, 0.1% for pitch and yaw. No time dependence is found and the overall mean errors are close to 0 with SD (systematic error on the overall population) within 0.1 mm (table). About the “no-tracking” scenario, a time dependence is found (figure), especially for left- right (Δy) translations. Up to 3' after starting the delivery no shifts larger than 2mm are observed, up to 5' they vary within a 3mm range. Then, major shifts appear mostly due to large (up to 9 mm) patient motion and time-related continuous shifts: after 10’ and 20' Δy was > 3mm in respectively 4 and 10 fractions out of 51.

motion were selected. The time-weighted average motion for each direction was calculated for all fractions of each patient. Thereafter, the mean intrafraction motion for each direction was calculated for both wERB and nERB groups. Planning CTs (pCTs) of 4 representative PCa patients treated at our institution were used to create virtual proton plans, using an extreme hypofractionated regimen of 4 fractions of 9.5 Gy(RBE), applying robust optimization using 5 mm setup and 3% range uncertainty to fulfill V100>95% for the clinical target volume (CTV). The calculated mean intrafraction motion for both groups was applied to the pCTs of those 4 patients and 44 synthetic CTs (sCTs) were created using deformable image registration in RayStation 6.99 as a surrogate to real time 4D calculations. Differences in CTV coverage and max dose to the organs at risk between the sCTs and pCTs were compared. Results Mean and standard deviation of intrafraction motion of both groups are listed in Table 1. The influence of intrafraction motion was negligible for CTV coverage which was maintained (V100>95%) in both groups. No detectable dose differences were found for the bladder. Posterior prostate motion resulted in a median max dose decrease to the anorectum of 61 cGy (wERB) and 65 cGy (nERB); inferior prostate motion resulted in a median max dose decrease of 53 cGy (wERB) and 57 cGy (nERB), respectively (Fig. 1A). A median max dose decrease to the anal canal of 0 cGy (wERB) and 158cGy (nERB) was found for posterior prostate motion and 92 cGy (wERB) and 184 cGy (nERB) for inferior prostate motion, respectively (Fig. 1B).

Conclusion Worst-case intrafraction prostate motion for both wERB and nERB groups did not negatively influence the