ESTRO 38 Abstract book

S537 ESTRO 38

retrospectively sorted 4D-MRI. The 3D-corrected images were considered as ground truth. All image corrections were derived from manufacturer-provided spherical harmonics coefficients. In this study, we included axial 4D- MRIs of three oligometastatic liver patients scanned on the MR-Linac (total: 10 imaging fractions). First, a step-and- shoot IMRT MR-linac plan (3x20 Gy, 10-13 beams) was created in Monaco 5.4 for each patient on the planning mid-position (midP)-CT. Next, each 4D-MRI was warped into a midP-MRI prior to applying the image corrections. The treatment plan was then adapted for each imaging fraction based on a rigid registration of the tumour mask in the planning midP-CT to the 3D-corrected midP-MRI (‘adapt-to-position’). Last, the planning midP-CT was deformably registered to each version of the daily midP- MRIs. The adapted plan was then re-calculated and evaluated using these daily midP-CTs. Results Visual inspection of the daily midP-MRIs in the axial orientation revealed mostly consistent contours between 3D- and 2D-corrected images. Uncorrected images featured local distortions of the external patient contour of up to 2 cm (fig 1). For the tumours, which were always positioned within 10 cm of the machine isocentre, 3D distortions were below 1.5 mm and hardly visible. Dose on the daily midP-CT showed local differences of up to 4 Gy/Fx compared to the original plan. Importantly, the daily dose differences induced by anatomical variations (3D–Planned) were much larger than the dose differences of the correction strategies (2D-3D, None-3D) for the PTV D95% and GTV D98% (fig 2). For the liver, the median increase in volume receiving less than 15 Gy was 31 cc (None) and 21 cc (2D) compared to the 3D- corrected images.

bronchi; n=1) or paracardial (PTV touching mediastinal or pericardial pleura; n=5). Risk-adapted fractionation was used to deliver 60Gy in 8 fx (n=19), 55Gy in 5 fx (n=4) or 60Gy in 12 fx using a GTV-PTV margin of either 3 mm (n=19) or 5 mm (n=5). For each fraction, daily MR-guided setup and on-table plan re-optimization based on PTV coverage and organ at risk (OAR) constraints was performed. Gated breath-hold delivery was performed under continuous MR-guidance. Of a total of 181 fractions, 167 were available for analysis. Benefits of daily plan re- optimization were studied by comparing 167 ‘predicted’ plans, which are the calculated baseline plans on the anatomy-of-the-day, with the re-optimized treatment plans, using Wilcoxon signed-rank test after exclusion of normality. Results In baseline plans, median PTV was 30.5 cc (range 4.2- 70.2), and 95% of PTV received a median dose of 60Gy (55.0-62.6) and a BED 10 of 105Gy (85.0-115.5 Gy). The mean interfractional volume change of clinician- contoured GTV and PTV were 0.4 cc (-3.2 – 6.8) and 0.7 cc (-3.3 – 11.7), respectively, with 75% of fractions showing a GTV/PTV variability ≤1cc when compared to baseline. Clinicians had chosen re-optimized plans for treatment in 91% of fractions; reasons included better PTV coverage (62%), OAR sparing (5%), a combination of PTV coverage / OAR sparing (12%), or because institutional protocol for some other tumor sites routinely selected this option (12%). PTV coverage by the prescription dose improved in re-optimized plans (PTV V100% 87.7% vs. 92.2%; p<0.01), leading to an increase in fractions fulfilling the PTV V100% ≥95% objective from 24.6% to 85.6%. A benefit for adaptation persisted when the 40 fractions with a GTV/PTV change ≥1cc from baseline were excluded from analysis (PTV V100% 87.7% vs. 91.6%; p<0.01). Median GTV and PTV doses, however, showed smaller improvements after re-optimization (GTV D50 69.0 vs. 69.3Gy, p<0.05; PTV D50 65.9 vs. 66.3Gy, p<0.01). Predefined OAR planning constraints were violated more often in predicted plans than in re-optimized plans (mean 0.73 vs. 0.53 violations per fx; p<0.05), and re-optimization allowed for better sparing of OARs in all cases when this was cited by the clinician as the reason for plan In central lung tumors treated using SMART, on-table plan adaptation improved PTV coverage, while avoiding excessive OAR doses. PO-0982 The dosimetric impact of geometric image distortions in slice-based 4D-MRI on the MR-linac M. Fast 1 , R. Keesman 1 , W. Van den Wollenberg 1 , C. Juan- Cruz 1 , T. Van de Lindt 1 , U. Van der Heide 1 , J. Sonke 1 1 Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective 4D-MRI is the modality of choice for guiding online plan adaptation of thoracic and abdominal lesions treated on the MR-linac. Like other MRI sequences, 4D-MRI is subject to scanner-specific geometric image distortions caused by gradient non-linearity and main magnetic field inhomogeneity. Specifically for research 4D sequences based on multi-slice 2D acquisitions, no 3D correction is currently available on the Elekta Unity MR-linac (Elekta AB, Stockholm, Sweden). Instead, only a 2D in-plane correction is applied leaving residual through-plane distortions unchanged. This study compares the dosimetric impact of (residual) geometric distortions on the online plan adaptation workflow of the MR-linac for liver SBRT. Material and Methods We investigated three different correction strategies for geometric image distortions on the Elekta Unity: A) None; B) 2D, in-plane distortion correction (current default); C) 3D, in-house developed full correction of the adaptation. Conclusion