S2233
Physics - Intra-fraction motion management and real-time adaptive radiotherapy
ESTRO 2026
Digital Poster Highlight 2630
To align with conventional margin design, we required the CTV to be fully contained within the PTV in 90% of the fractions [1]. The clinical acceptance frequency for performing a secondary fast rigid Adapt To Position (ATP) was also incorporated and chosen to be ≤ 10 %.A customised 3D U-Net was trained to predict dose distributions using segmentations of the CTV, bladder and rectum. The clinical TPS doses served as ground truth. Data were split into 120/24/62 fractions for training, validation and testing respectively. A tailored loss function combining mean squared errors and isodose volume Dice scores ensures both voxel wise accuracy and correct conformity. The test set was evaluated using common treatment planning indeces to confirm clinical similarity.For dose accumulation, a PTV with reduced margins was generated. CTV positions were sampled every 5 seconds from motion traces. The CTV was shifted within the grid of the predicted dose distribution given the static cloud approximation. MU delivery data were retrieved from the machine logfiles. The dose accumulation assumed MU weighted delivery of the entire dose distribution at each time sample. Results:
Optimized segment-specific PTV margins for dose- escalation radiotherapy in rectal cancer based on the MR-Linac Wang Yang, Lei Yu, Yiwen Hu, Hui Zhang, Yajie Chen, Menglong Zhou, Yaqi Wang, Zhen Zhang, Fan Xia Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China Purpose/Objective: Dose escalation in rectal radiotherapy represents a promising approach for improving complete response rates and potentially facilitating organ preservation. However, the optimal planning target volume (PTV) margin for rectal lesions has yet to be established. The recent adoption of MR-guided radiotherapy provides a unique opportunity for margin analysis owing to its capability for high-resolution and real-time visualization. Nevertheless, the prolonged treatment duration required by this technique may introduce additional uncertainties in intrafraction motion. The purpose of this study is to characterize rectal motion uncertainties, establish optimized segment-specific PTV margins, and perform dosimetric validation using MR-Linac data to enhance treatment accuracy in rectal cancer dose escalation. Material/Methods: This study analyzed 22 advanced rectal cancer patients treated with neoadjuvant radiotherapy on a 1.5T MR- Linac. Pre-, intra-, and post-treatment MR images were obtained for each fraction, yielding 334 scans from 113 fractions for analysis. The rectum was segmented (upper, middle, lower) per the distance from the anal verge. Inter- and intrafractional margins were established through isotropic expansion of the segmented clinical target volume (CTV), with margin adequacy defined as covering 95% of the CTV volume in 90% of all fractions. Dosimetric validation was conducted in 15 patients (10 internal, 5 external validation), comparing the proposed segment-specific margins against uniform margins from published literature for target coverage and organs at risk (OARs) Rectal motion exhibited segment-dependent variability, with the largest margin requirements observed in the upper rectum and the smallest in the lower rectum. The determined interfractional margins were 13 mm for the upper, 7 mm for the middle, and 6.5 mm for the lower rectum. Intrafractional analysis further demonstrated that treatment duration significantly influenced margin requirements: for treatments under 30 minutes, margins of 7 mm for the upper rectum and 3 mm for the middle and lower rectum proved sufficient, while longer treatment durations required moderately increased margins. sparing. Results:
The reduced PTV margins that resulted in an ATP frequency of < 10 % were (4, 5, 5) mm coupled with an ATP threshold of (3, 3, 3) mm, both (L-R, A-P, S-I).The resulting CTV D98% based on accumulated doses for all simulated fractions, seen in Figure 1, show that less than 0.5 % of the fractions yield a D98% lower that the clinical limit at 95% of the prescribed dose. No fraction resulted in a D98% higher than 105%. The lower 10th percentile of D98% is far from the clinically accepted 95% of the prescribed dose Conclusion: This work confirms the dosimetric adequacy of the reduced motion-based margins. Furthermore, it clearly shows the potential for further margin reductions. References: [1] van Herk et al, "Inclusion of geometric uncertainties in treatment plan evaluation", IJROBP 2002 Apr 1;52 Keywords: Dose accumulation prediction
Made with FlippingBook - Share PDF online