Session Item

Online MR-guided radiotherapy (MRgRT) enables optimal plan adaptation with respect to the patient’s daily anatomy, allowing for improved target coverage and organ at risk (OAR) sparing. Deformable dose accumulation (DDA) promises precise evaluation of organ-specific accumulated doses. However, various DDA algorithms and implementations are available, making QA, inter-institutional comparison and recommendations for clinical use of DDA  challenging. The aim of this multi-institutional analysis was to compare different software solutions and implementations of DDA and investigate differences in resulting dose metrics.

Six institutes participated in this DDA analysis. Five data sets of patients which were treated with online adaptive stereotactic MRgRT were included: (I) cervix (5 x 7 Gy), (II) liver (5 x 7 Gy), (III) lymph node (5 x 6 Gy) and (VI,V) two prostate (5 x 7.5 Gy) cases. The  T2w MRI for each fraction was re-contoured offline by an experienced radiation oncologist. Each institute performed a retrospective DDA using the software implementation available at their respective centre. First, a deformable image registration (DIR) was carried out, in which the MRIs of each fraction were deformably registered to a reference MRI (fraction 1). Three institutes used hybrid intensity/structure (A,B,C), one institute used contour (D), one intensity (E), and one normalised gradient fields/structure (F) based DIR algorithms. DIR quality was evaluated using dice similarity coefficients (DSC). Then, DDA was performed accordingly. Five institutes used direct dose mapping (A-E), one the energy mass transfer method (F). Resulting accumulated dose distributions were analysed with respect to the treating institution’s clinical dosimetric constraints (CDCs).

Analysis of DSC indicates better DIR quality for the hybrid intensity/structure-based algorithms especially for contours that were used as guidance. The DDA dose-volume histograms (DVH) per case are shown in figure 1. In general, good agreement was found between the different DDA approaches. The dosimetric results are summarised in table 1. For case (I), rectum D2ccm showed the largest range in OAR dose metrics with 2.8 Gy.  Duodenum with Dmax 7.1 Gy showed the largest OAR CDC difference for the case (II).  For case (III), D0.5ccm for rectum resulted in the largest OAR CDC difference with 4.0 Gy. The greatest CDC OAR difference was observed for bladder V28Gy with 10.2 Gy and 7.6Gy for the cases (VI,V), respectively.

Our focus in this analysis was the evaluation of the CDCs for different implementations of DDA. The results demonstrated good agreement of the CDCs, with a few exceptions, considering all the different approaches. This may have significant implications for potential future interventions or plan adaptations based on these metrics.