S2241
Physics - Intra-fraction motion management and real-time adaptive radiotherapy
ESTRO 2026
CBCT registration. The total mean 3D vector for all fractions was 4.1±5.9mm. Conclusion: This first-in-patient DIBH-SBRT for lung lesions confirms the technical and clinical feasibility of markerless using ExacTrac Dynamic SGRT with stereoscopic x-rays in combination with Correlation Objects for lesion localization, positioning and intrafraction motion monitoring. The workflow was efficient and well tolerated, supporting further prospective validation in a larger patient cohort. Keywords: Markerless lung SBRT, Stereoscopic X-rays, DIBH Digital Poster 2936 Particle therapy with AI-powered in vivo dose verification and tumor tracking Luana Testa 1,2 , Giuseppe Battistoni 3 , Alberto Burattini 4,5 , Marina Carruezzo 5,6 , Yunsheng Dong 3 , Gaia Franciosini 5,6 , Marco Garbini 2 , Laura Frassi 1 , Marco Magi 5,6 , Michela Marafini 2,6 , Ilaria Mattei 3 , Silvia Muraro 3 , Flaminia Quattrini 1,6 , Alessio Sarti 5,6 , Angelo Schiavi 5,6 , Marco Toppi 5,6 , Giacomo Traini 6 , Arianna Vannucci 2,4 , Vincenzo Patera 5,6 1 Department of Physics, Sapienza University of Rome, Rome, Italy. 2 Department of Physics, ”Enrico Fermi” Historical Museum of Physics and Study Research Center, Rome, Italy. 3 Milan Section, National Institute of Nuclear Physics (INFN), Milan, Italy. 4 Specialty School in Medical Physics, Sapienza University of Rome, Rome, Italy. 5 Department of Basic and Applied Sciences for Engineering, Sapienza University of Rome, Rome, Italy. 6 Roma I Section, National Institute of Nuclear Physics (INFN), Rome, Italy Purpose/Objective: Particle therapy (PT) is an advanced form of radiotherapy that exploits the unique energy deposition profile of charged particles, protons and carbon ions, to deliver highly conformal dose distributions, maximizing tumor control while sparing healthy tissues [1]. However, its clinical precision remains limited by range uncertainties and patient motion, particularly in lung cancer, where respiratory displacements may exceed 20 mm. These variations compromise dose accuracy, and no robust system currently enables real-time monitoring. This study aims to develop and validate a technology for real- time in vivo dose verification and tumor tracking, capable of monitoring anatomical changes and beam range during irradiation to support truly adaptive,
The proposed strategy exploits the interactions between a three-dimensional matrix of scintillating fibers and the secondary particles generated by the therapeutic beam during patient irradiation, establishing a direct correlation with the dose deposited in vivo. The detector captures the full flux of secondary radiation (photons, neutrons, and charged fragments). The research workflow involves the design and experimental characterization of a compact, high- granularity scintillating-fiber detector with SPAD-based (Single-Photon Avalanche Diode) optical readout [2,3], the implementation of advanced Monte Carlo simulations (FLUKA [4] and GPU-based codes like FRED [5]) to generate realistic detector images for Artificial Intelligence (AI) training, and the development of dedicated deep-learning models, including a Convolutional Neural Network (CNN), to extract in real time key clinical parameters such as beam range, dose distribution, and tumor motion. Results:
A proof-of-principle study performed using FLUKA simulations demonstrated the feasibility of localizing the irradiation position through CNN analysis of simulated detector images. A 10 cm radius water sphere irradiated with a 190 MeV, 10E7 proton pencil beam was reconstructed by the CNN, trained on images generated from the simulated energy deposited by secondary particles in the scintillating fibers. After appropriate training, the network was able to determine the spatial position of the irradiated region with an accuracy of about 3 mm, confirming the
motion-resilient PT. Material/Methods:
Made with FlippingBook - Share PDF online