S2111
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
reproducibility. Material/Methods: We retrospectively analyzed 36 patients who received lung SBRT using a BH technique over a one - year period on a single linear accelerator. In total, 142 pre - treatment CBCT scans were reviewed. The details of the treatment delivery procedure are provided in a prior publication1. To quantify BH reproducibility, we measured the shifts between alignment to adjacent vertebra and alignment to the gross tumor volume (GTV) for each fraction. For each treatment fraction, we calculated the mean and SD of these shifts in the superior–inferior (SI), anterior–posterior (AP), and left– right (LR) axes, as well as the composite “total shift” (computed as the root sum square of the three directional shifts). Spearman’s rank correlation tests were conducted to examine associations between total shift magnitude and patient demographic/clinical variables (age, sex, disease stage, body mass index, tumor laterality) and the time interval between simulation and treatment. In addition, we analyzed the correlation of total shift magnitudes across fractions to evaluate BH inter-fraction reproducibility. Results: The mean ± SD shifts (mm) in the SI, AP and LR in directions were 4.7 ± 3.7, 2.4 ± 3.0 and 1.3 ± 1.4 for fraction #1; 5.2 ± 5.0, 2.6 ± 3.5 and 1.5 ± 2.0 for fraction #2; 5.3 ± 4.9, 3.0 ± 4.0 and 1.2 ± 1.0 for fraction #3; 4.7 ± 4.3, 2.6 ± 2.9 and 1.2 ± 1.3 for fraction #4. Composite total shifts (mm) were 5.9 ± 4.3, 6.6 ± 5.8, 6.5 ± 6.0 and 6.1 ± 4.8 for fractions 1 to 4 respectively.No strong correlations were found between composite total shifts (i.e. BH reproducibility) and age, sex, stage, BMI, tumor laterality (|r| < 0.4), or the simulation - to - treatment interval (|r| < 0.2). However, total shifts across fractions correlated moderately to strongly (r ≥ 0.4).
Metrics shown include rotational MAE in degrees (top left), translational MAE in mm (top right), BM barycenter shift in mm (middle), DICE in % (bottom left), and HD in mm (bottom right).
Conclusion: Quantifying expert variability provides a probabilistic benchmark that reflects realistic clinical practice. An AI-driven CB algorithm consistently reached expert- level accuracy, reduced operator dependence, and improved the performance of less-experienced users. This framework enables clinically meaningful validation of automatic CT–MRI registration methods, supporting safer and more standardized integration into brain SRS workflows. Keywords: SRS, Multimodality Imaging, Algorithm Validation Digital Poster 447 Breath-Hold Reproducibility in Lung SBRT: from Simulation to Treatment Xiaochun Wang, Sandun Hewage, Christopher Peeler, Julianne Pollard-Larkin, Ramaswamy Sadagopan, Paige Nitsch, Suman Shrestha, George Zhao Radiation Physics, MD Anderson Cancer Center, Houston, USA Purpose/Objective: To evaluate the reproducibility of RPM - guided breath hold (BH) from simulation through treatment in lung SBRT, and to investigate the factors that influence BH
Made with FlippingBook - Share PDF online