ESTRO 2025 - Abstract Book

S3039

Physics - Image acquisition and processing

ESTRO 2025

compare clinical and MR-only workflow, both were compared to a “superior” ground truth, consisting of CT-based dose calculations and manual delineated OARs on the T2w scan. Recently defined guidelines were used, based on the DAHANCA and RTOG CT guidelines and adapted for MR. 3 Figure 1 shows examples of the delineations. OARs and images were deformably registered to the CT’s frame of reference. The clinical plan was recalculated on the synCT and assessed in the OARs. The registrations and manual delineations were manually checked by an experienced radiation oncologist. The absolute dose differences (Dmean for parallel and D(2%) for serial structures) to the ground truth were compared between the clinical and automated MR-only workflows using a Wilcoxon signed rank test.

Results: Figure 2 shows the differences between the clinical workflow and MR-only workflow compared to the ground truth. On average these differences were -1.39±3.01 Gy and 0.34±2.19 Gy, respectively). Three out of twelve OARs had significantly larger dose errors for the clinical workflow compared to the MR-only workflow (p<0.05).

Conclusion: The dosimetric results from the fully automatic MR-only workflow are closer to the ground truth than the current clinical workflow. Therefore, the data suggests that the automated MR-only workflow outperforms the current