S2110
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
Institute, Villejuif-Paris, France. 5 Department of Physics, University of Milan, Milan, Italy. 6 School of Medical Pysics, University of Padua, Padua, Italy. 7 Radiation Oncology, Fondazione IRCCS Istituto Nazionale Dei Tumori, Milan, Italy. 8 Health epartment, Department of Neurosurgery, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy. 9 Radiation Oncology, Fondazione IRCCS San Gerardo dei Tintori, Monza, Italy Purpose/Objective: Accurate CT–MRI registration is critical for stereotactic radiosurgery (SRS) of brain metastases (BM), since treatment planning depends on CT but lesion visualization relies on MRI. Despite being considered the clinical reference, manual registration by experts shows measurable variability, which is particularly relevant given the submillimetric margins used in modern SRS. Most validation studies for automatic registration rely on phantoms or single-expert references, failing to capture clinical complexity. This study aimed to quantify expert variability, establish a probabilistic gold standard (GS), and use it to benchmark automatic registration algorithms. Material/Methods: Twenty CT–MRI pairs (39 BM) were registered twice by six operators (n=240). Variability was assessed by bootstrap resampling to estimate 99% confidence intervals (CIs) for rotational and translational mean absolute error (MAE), BM barycenter shift, DICE, and HD. For each case, the GS was defined as the median translational and rotational components across experts. Deviations from the GS were bootstrapped to derive expert-to-GS variability, used as an acceptance range. Three mutual-information (MI)-based methods were tested: standard MI, skull-box MI, and contour- based (CB) MI (manual or AI-generated contours). Less-experienced operators were also evaluated. Results: Experts showed low but non-negligible variability (99% CI upper bounds: 0.31° rotation, 0.34 mm translation, 0.82 mm barycenter shift). Barycentre shifts exceeding 1 mm occurred in ~30% of BM, with deviations up to 2.8 mm. Only the CB algorithm achieved median values within expert variability for all metrics except HD (Figure 1), with no significant differences between manual and AI-generated contours (P>.05). Less- experts showed larger deviations, markedly reduced when starting from CB-aligned datasets.Figure 1: Box plots of the different registration modalities compared to the expert-derived GS. From left to right, the reported outputs are from experts, first round (R1) and second round (R2) of less-experts, standard mutual-information (MI) algorithm, skull box-based algorithm, manual-contoured brain-based (manual- brain CB) algorithm, and AI-driven automatic- contoured brain-based (auto-brain CB) algorithm.
(Gy)0.100.020.23V8Gy (%)1.170.302.50Normal TissueDmax0.03cc (Gy)0.340.021.09Monitor UnitsMU Difference235201308Workflow MetricsAverageMinMaxImage Selection and Lattice Generation8.54.012.0Reference Plan Creation21.518.029.0Adaptive Contouring4.12.55.6Plan Optimization and Calculation6.85.28.7Total Patient time166.4150.7180.2
Conclusion: This in-silico study demonstrates the feasibility of same-day, simulation-free SF ² -ART. The workflow maintained dosimetric quality, respected OAR limits, and proved technically deliverable without simulation CT. Integration of automation, adaptive optimization, and CBCT enables treatment initiation in a single session. This approach may reduce time-to-treatment from over a week to less than four hours and supports prospective clinical implementation for urgent or symptomatic patients. References: Prezado Y et al. Phys Med Biol. 2024;69(10).Duriseti S et al. Adv Radiat Oncol. 2021;6(3):100639.Schiff JP et al. Adv Radiat Oncol. 2023;8(1):101091. Keywords: Adaptive radiotherapy;spatially fractionated SBRT Digital Poster Highlight 205 Expert variability as a benchmark for validating automatic CT–MRI registration in brain Stereotactic Radiosurgery Valeria Faccenda 1,2 , Denis Panizza 1,2 , Valentina Pinzi 3,4 , Stefano Carminati 5 , Riccardo Ray Colciago 2 , Stefano Alberto Mascellani 2 , Bianca Bordigoni 6 , Ilenia Manno 2 , Matteo Mombelli 2 , Vanessa Eleonora Pierini 7 , Alicia Rejas Mateo 7 , Elena De Martin 8 , Elena De Ponti 1 , Stefano Arcangeli 2,9 1 Medical Physics, Fondazione IRCCS San Gerardo dei Tintori, Monza, Italy. 2 School of Medicine and Surgery, University of Milan Bicocca, Milan, Italy. 3 Radiotehrapy Unit, Department of Neurosurgery, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy. 4 Radiation Oncology, Gustave Roussy Cancer Center
Made with FlippingBook - Share PDF online