ESTRO 2022 - Abstract Book

S836

Abstract book

ESTRO 2022

Results (1) Model performance was 1.7 +/- 1.1 mm (mean prediction error, validation set). Mean interfractional amplitude variability was 3.4 +/- 2.4 mm. 4-fold cross validation obtained similar mean prediction errors, indicating model reliability. (2) Results of ITV re-definition after the first fraction are summarized in Table 1. The test set prediction error was in the range of 0.3 to 4.2 mm and thus greater than for the validation set. However, the test set interfractional amplitude variability was significantly higher compared to the validation set. Despite reduction of initial ITV size, the resulting adapted ITV would have achieved sufficient tumor motion coverage during dose delivery when implemented in the clinics.

Conclusion Our results support the assumption that 4DCT-based ITV definition may lead to unfavorable ITVs. However, this mainly occurs in cases of adverse interplay of breathing-sensitive tumor motion and interfractional variability of patient breathing. The proposed approach of ITV re-definition reveals potential of significant ITV volume reduction and normal tissue sparing.

OC-0944 Adaptive fractionation at the MR-Linac based on a dynamic programming approach

Y.S. Pérez Haas 1 , R. Ludwig 1 , R. Dal Bello 1 , L. Wenhong 1 , S. Tanadini-Lang 1 , J. Unkelbach 1

1 University hospital Zurich, Radiation Oncology, Zürich, Switzerland

Purpose or Objective Inter-fraction motion of tumors and dose-limiting OARs can be visualized using MR guidance. Whereas standard treatments deliver the same dose in each fraction, adaptive fractionation (AF) is an approach to exploit inter-fraction motion by increasing the dose on days when the distance of tumor and OAR is large and decreasing the dose on unfavorable days. We develop an algorithm for AF and evaluate the concept for former patients treated at the MR-Linac for abdominal tumors in 5 fractions. Materials and Methods Given daily MR scans and adaptive treatment plans, inter-fractional changes in fraction t are quantified by sparing factors δ t , which are defined as the ratio of dose delivered to the OAR (D 1cc ) and the tumor (D 95% ). The key problem of AF is to decide on the dose to deliver in fraction t , given today's δ t and the dose delivered in previous fractions, but with unknown future δ s. This problem can be formulated as a Markov decision problem and solved with a dynamic programming algorithm. The algorithm assumes a normal distribution over δ with mean and variance estimated from previously observed patient- specific δ s and a population based prior for the variance. Based on the distribution, the algorithm computes optimal doses that maximize the expected tumor BED 10 while staying below the maximum BED 3 of 90 Gy in the OAR (30 Gy in 5 fractions). To evaluate the algorithm, 10 MR-Linac patients were exported in whom tumor dose was compromised due to proximity of bowel, stomach, duodenum or heart. Moreover, 1000 synthetic patients with similar δ distribution have been sampled. AF was compared to a standard plan which delivers 6 Gy to the OAR in each fraction to reach exactly the 90 Gy BED 3 constraint. Results In 7 of the 10 patients, AF was equal or increased the tumor BED 10 , on average by 0.72 Gy (1.2%). Figure 1 shows the δ 's and the respective doses based on AF for all patients. The biggest increase of 13.1 Gy (18.4%) was achieved for patient 2 with δ =[0.88,0.8,0.58,0.86,0.77] where AF delivered tumor doses of [4.4,8.7,17,0.9,2.1] and thereby exploited the low δ in fraction 3. Figure 2a) illustrates the optimal policy for fraction 3, i.e. optimal dose to deliver as a function of δ . For the 1000 generated patients, 84.1% of plans were equal or superior to the standard plan with a mean improvement of 1.01 Gy (1.5%) BED 10 . A distribution of the differences is shown in Figure 2b). With the most extreme plans having a benefit of up to 12 Gy BED 10 .