ESTRO 2020 Abstract Book

S891 ESTRO 2020

PO-1629 A quantitative evaluation of deformable image registration based on Halcyon MVCBCT images Y. Huang 1 , H. Wu 1 , W. Wang 1 , Y. Zhang 1 1 Key Laboratory of Carcinogenesis and Translational Research Ministry of Education/Beijing- Department of Radiation Oncology- Peking University Cancer Hospital & Institute, Department of Radiotherapy, Beijing, China Purpose or Objective The incorporable imaging dose and fractional guidance of Halcyon MV cone beam CT images (MVCBCT) inspires its potential application on adaptive radiotherapy (ART) using deformable-image-registration-based (DIR). As a precondition, however, the accuracy dependence on the deformation magnitudes and image modalities remains unknown for Halcyon, which is the aim of this study. Material and Methods Planning CT images of an anthropomorphic Head phantom were aligned rigidly with MVCBCT and re-sampled to achieve the same resolution, denoted as pCT. MVCBCT was warped with twenty simulated pre-known virtual deformation fields (VFi , i= 1 to 20) with increasing deformation magnitudes, yielding warped CBCT (wCBCT). The wCBCT was registered to pCT and MVCBCT respectively (Multi- and Uni-modality DIR), generating deformation vector fields Vi and Vi' (i= 1 to 20). Vi and Vi' were compared with VFi respectively to assess the DIR accuracy geometrically. In addition, Vi, VFi and Vi' were applied to pCT, generating three sets of deformed CT: dCTi, wCTi and dCTi' respectively. The Hounsfield Unit (HU) on these virtual CT images were compared as dosimetric impacts. Results The impact of deformation magnitudes and imaging modalities on the geometric and HU accuracy of DIR has been investigated based on the new Halcyon MVCBCT system. As deformation magnitude increases, the mean errors of vector displacement continued to deteriorate. Uni-modality DIR consistently outperformed multi- modality DIR. For deformation magnitudes between 2.82 mm to 7.71 mm, the errors of uni-modality DIR were 1.16 mm~1.73 mm smaller than that of multi-modality (p=0.0001, Wilcoxon signed rank test). Considering direct dose comparison is rather plan-dependent and patient- specific and hence is ungeneralizable, HU was used as a surrogate dosimetric indicator of DIR accuracy in this work. By applying multi/uni-modality DIR, the maximum HU deviations could be reduced from 70.8 HU /208 HU to 12 HU/ 47 HU for signed/absolute HU errors respectively. Conclusion Although it is a preferable condition that imaging dose of Halcyon MVCBCT can be incorporated into the treatment dose, DIR-based ART utilizing the noisy MVCBCT images should be applied with caution. PO-1630 Accurate 3D-rotation correction for stereotactic intracranial treatments without a 6D couch S. Kwa 1 , A. Méndez Romero 1 , M. Hoogeman 1 1 Erasmus MC Cancer Institute - Rotterdam, Department of Radiation Oncology, Rotterdam, The Netherlands Purpose or Objective Rotation correction is considered a prerequisite for accurate stereotactic treatment (SRT) of elongated or simultaneously treated multiple lesions in the brain. However, a 6D couch is not always available. The purpose of this work is to provide an effective rotation correction in fractionated SRT in the absence of a 6D couch Material and Methods Fractionated SRT was performed on a standard linac for 18 patients with benign intracranial tumors (i.e. single lesions only) in N = 28–30 fractions. The patients were immobilized with a double shell mask (Double Shell Positioning System; DSPS - MacroMedics). Inter-fraction translational and

rotational errors were derived from a 6D grey-value registration of the daily acquired CBCT scan with the planning CT scan. Only translational errors were corrected prior to treatment. For each patient the average rotational error during the first 3 fractions (R 1–3 ), the remaining fractions (R 4–N ), and all fractions (R all ) was calculated around the LR, CC, and AP axis. We proposed the following offline rotation correction (ORC) protocol to correct for systematic rotations: The dose distribution will be rotated (not the patient/couch) and R 1–3 is used to predict the systematic rotations of the remaining fractions R 4–N . If R 1–3 is higher than 0.5 o or 1.0 o , the planning CT with delineated structures is rotated in opposite direction of R 1–3 and used to adapt the treatment plan for fractions 4 to N. The ORC protocol can be simulated for each patient by subtracting R 1–3 from the daily rotations in fractions 4 to N. R all,ORC (around 3 axes) is the residual rotational error after the applied ORC. ORC was also simulated for N=3. Results The systematic rotational errors R all were 2.2 o or less for all patients. The population standard deviations Σ were 0.5 o , 0.5 o , 1.0 o for the LR, CC, and AP rotation axes, respectively. Figure 1 shows that R 1–3 was highly correlated with R 4–N : the Pearson correlation coefficient was 0.88, 0.90, and 0.97 for the LR, CC, and AP rotation axes, respectively (error bars denote the standard deviation). For ORC with an action level of 0.5 o , Σ reduced to 0.3 o , 0.2 o , 0.3 o (for LR, CC, AP) and R all,ORC was 0.5 o or less for all patients (Fig. 2; plan adaptation was required for 13/18 patients). For an action level of 1.0 o , these rotations were slightly higher: Σ = 0.4 o , 0.2 o , 0.4 o (for LR, CC, AP; plan adaptation required for 8/19 patients) and R all,ORC ≤ 0.9 o . When using the rotation of the first fraction only (R 1 ) to predict the average rotation of fractions 2, 3, and 4 (R 2–4 ), ORC with an action level of 1.0 o reduced Σ from 0.6 o , 0.6 o , 1.0 o (LR, CC, AP; R 2–4 ≤ 2.3 o ; without ORC) to 0.5 o , 0.3 o , 0.5 o (R 2–4,ORC ≤ 1.3 o ; with ORC).

Fig. 1