ESTRO 2022 - Abstract Book

S845

Abstract book

ESTRO 2022

Conclusion Significant deviation from the reference plan dose for rectum was observed with both systems due to large range of anatomical changes exhibited between planning and treatment sessions. MR-guided adaptation was able to deliver a lower than reference plan dose to the rectum in the high dose region more frequently than CBCT-guided 3DOF, and these patients experienced a lower incidence of G2+ GI acute toxicity. Follow up is ongoing and will assess the clinical efficacy of these two treatment delivery systems.

OC-0953 Dosimetric validation of a hybrid DIR algorithm for MR-Linac dose accumulation

V. Malkov 1 , J.D. Winter 1,2 , V. Kong 1,2 , W. Li 1 , J. Dang 1 , I. Navarro 1 , J. Padayachee 1,2 , P. Chung 1,2 , T. Tadic 1,2

1 Princess Margaret Cancer Centre, Radiation Medicine Program, Toronto, Canada; 2 University of Toronto, Department of Radiation Oncology, Toronto, Canada Purpose or Objective The MR-linac adapts treatment plans to accommodate inter-fraction anatomical changes. Patients will experience further intra-fraction motion which will impact the final delivered dose. We aim to assess the performance of a hybrid intensity- and structure-based DIR algorithm for inter- and intra-fraction prostate MR images using geometric and dosimetric assessments. Materials and Methods We generated manual contours of the bladder, rectum, and CTV on 3D T2 MR images of 25 patients with prostate cancer treated with 30Gy/5 SBRT. In addition to the reference MR scan, three MR scans were obtained for each patient per fraction: session start (adapt), end of plan adaptation (verify), and during beam delivery (beam-on). We contoured 400 MR images. We recomputed adapt doses on the verify and beam-on MR scans. We performed DIRs for each patient between the reference MR and all fraction images, as well as within fraction adapt-to-verify (A-to-V) and adapt-to -beam-on (A-to- B). We use a hybrid DIR algorithm with two strategies: image intensity alone, and image intensity with controlling structures. The latter uses Chamfer matching to optimize the registration of the bladder, rectum, and CTV ROIs. For a subset of patients, we included controlling point-of-interest (POIs) in the DIR. We automatically generate the POIs based on an optimized mesh representation of the bladder adapted across all MR images. We compare manual and mapped contours using max distance to agreement (mDTA), dice similarity coefficient (DSC), and difference of DVHs (dDVH). Results In Fig. 1 we compare DSC between manual and mapped structures for the A-to-V and A-to-B DIRs using intensity- only and combined intensity-structure strategies. Use of the controlling ROIs improved the DSC and reduced variance. For the A-to-B intensity-only DIR, mDTA was 0.34±0.13 cm (CTV), 2.1±1.2 cm (bladder), and 0.9±0.6 cm (rectum). For the intensity-structure DIR, the values were 0.14±0.06 cm, 0.25±0.13 cm, and 0.3±0.6 cm, respectively. In Fig. 2 we present the dDVH between the manual and A-to-B mapped contours. For 14 patients, use of the controlling ROIs notably reduces dose differences across the entire dDVH. For 9 patients, the combined intensity-structure DIR resulted in worse ROI alignment (mDTA of 0.22±0.11 cm, 2.6±2.8 cm, 0.8±0.7 cm for A-to-B). By using controlling POIs, we were able to improve these DIRs (mDTA of 0.20±0.09 cm, 0.27±0.14 cm, 0.7±0.7 cm). The DIR between the reference and beam-on