ESTRO 2024 - Abstract Book

S4269

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

specific information, we hypothesize that it enables a reduction of healthy tissue dose whilst maintaining CTV coverage with respect to the standard clinical workflow.

Material/Methods:

14 rectal cancer patients that were treated with online adaptive radiotherapy with 5 x 5Gy on a 1.5T MR-Linac were retrospectively selected. Each patient had a pre- and a post-treatment scan for each fraction. On all pre- and post treatment scans, the mesorectum CTV was delineated. The PTV-margins for the first 4 fractions were 0 mm in all directions, except for 5 mm in cranial-caudal (cc) direction (=PTV0) because of substantial inter-observer variation in the cranial and caudal boundaries of the CTV delineations. For fractions 1-4, a daily treatment plan with PTV0 as target was generated on the pre-scan and recalculated on the post-scan as an estimate of the actual delivered dose. The recalculated doses of the first 4 fractions were deformed using a B-spline deformable image registration (DIR) algorithm and accumulated on the pre-scan of fraction 5 as a base dose. To evaluate the DIR accuracy, dice coefficient (DC) and 95% Hausdorff distance (HD95) of the deformed and target CTV were calculated. For fraction 5, a plan was optimized such that, accumulated with the base dose, in PTV0 the total prescribed dose was reached. The plan was recalculated on the post-scan of fraction 5 and dose accumulation of the complete treatment was performed on the post-scan of the fraction 5. Accumulated dose was compared with the accumulated dose from conventional online adaptive radiotherapy using the clinical PTV margins of 8 mm anterior and 5 mm otherwise (=PTVclin). This workflow is shown in Fig. 1. Accumulated target coverage (V 95% >99%, V 100% ) and near max dose (D 1% <107%) in the CTV were evaluated. The dose sparing of the healthy tissue was evaluated via D 50% in a ring of 1 cm around the PTVclin on the post scan of fraction 5. Significance of the dose differences was tested with the Wilcoxon signed-rank test with an α of 0.05. The average DC and HD95 of all fractions and patients were 0.98 and 1.2 mm, respectively, concluding a high accuracy of the DIR. Online dose adaptive RT with PTV0 led to sufficient target coverage (V 95% >99%) for all patients with no significant difference compared to conventional online adaptive RT using PTVclin (p = 0.29). The V 100% was not significantly different (p = 0.14) and the mean difference of the V 100% was 4.8%. A reason that a sufficient target coverage is accomplished with 0 mm margin might be the dose inhomogeneities. To compensate the missing dose after 4 fractions, larger inhomogeneities (D 1% >107% of 5Gy, physical dose) than in clinical practice were necessary. Dose compensation was primarily required around the surface of the PTV0 leading to a larger dose spill in regions that are prone to underdosage and hence compensating for intra-fraction motion in the last fraction. Accumulated D 1% was on average 0.5 Gy higher (p < 0.005) when using PTV0 instead of the conventional adaptive technique. However, only one patient had a D 1% >107% (27.8Gy) when using PTV0. The dose to the healthy tissue was lower with PTV0 in all cases. The average D 50% for the ring around the PTVclin was 21.9 Gy when using PTVclin as target. This value was reduced on average by 2.1 Gy for PTV0 with an interquartile range of the reduction of 0.4 Gy. The DVH of the patient with the medium dose reduction of D 50% is shown in Fig. 2. Results: