S3044
Invited Speaker
ESTRO 2026
complex solution spaces beyond classical strategies. Within this framework, classical optimization variables are mapped onto quantum bits (qubits), and the objective function is encoded into a Hamiltonian, whose ground state encodes a set of optimal treatment plans. The lecture will guide the audience through two main application scenarios in radiotherapy inverse planning. First, quadratic optimization problems will be discussed. Second, more complex non-polynomial and non-convex problems will be addressed. Given that current implementations operate in the Noisy Intermediate-Scale Quantum (NISQ) era, the lecture will also critically discuss practical limitations, including the restricted number of qubits, and limited connectivity. Hybrid classical–quantum and quantum- inspired approaches will be introduced, with examples such as Tree Tensor Networks (TTN), demonstrating the feasibility of the approach through preliminary results. The lecture aims to provide attendees with a conceptual understanding of how quantum computing may complement and extend current radiotherapy optimization strategies. 5452 Direct hit technique for liver oligometastases Mateusz Edward Bilski Radiotherapy Department, Affidea Nu-med Center of Oncological Diagnostics and Therapy, Zamo ść , Poland. Clinics of Radiotherapy, Medical University of Lublin, Lublin, Poland To evaluate clinical outcomes and dosimetric characteristics of computed tomography-guided interstitial high-dose-rate brachytherapy (CT-HDR-BT; interventional radiotherapy) for liver oligometastases across different clinical scenarios, and to compare its performance with stereotactic body radiotherapy (SBRT). Data from multiple real-world cohorts of patients with liver oligometastases treated with CT-guided HDR brachytherapy were analyzed, including colorectal cancer (CRC) and breast cancer (BC) (1–4). Patients with de novo, repeat, and induced oligometastatic disease, defined according to EORTC/ESTRO classification, were included. Treatment consisted of percutaneous CT-guided catheter placement followed by single-fraction HDR irradiation. Clinical endpoints included overall survival (OS), progression-free survival (PFS), local control, and acute and late toxicity. Prognostic and predictive factors were evaluated using multivariable analyses. In parallel, dosimetric comparisons between HDR brachytherapy and stereotactic body radiotherapy (SBRT; linac- and
principles, with a focus on their relevance to radiobiological applications. Building on this foundation, the talk will present model classes and use cases, spanning tasks such as radiosensitivity prediction, multimodal outcome modelling, and more advanced AI systems including emerging agent-based approaches. A central part of the lecture will address recurring challenges in radiobiology-driven AI, including limited cohort sizes, irregular and sparse sampling, endpoint uncertainty, and the difficulty of generating biologically meaningful labels. Potential mitigation strategies will be discussed, including the use of prior biological knowledge, feature binning, and knowledge- guided machine learning approaches that constrain or inform data-driven models using mechanistic insight. The lecture will conclude with an outlook on future directions, including agentic AI, applications in advanced radiotherapy delivery, and biologically informed personalized prediction, highlighting both the opportunities and the methodological rigor required for clinically and scientifically meaningful progress. 5451 Application in RT Samuele Cavinato Medical Physics Department, Veneto Institute of Oncology IOV-IRCCS, Padua, Italy As the physics community has just celebrated 100 years of quantum mechanics through the International Year of Quantum Science and Technology, quantum computing (QC) is emerging as a potentially transformative paradigm for radiation oncology. This teaching lecture aims to provide a structured introduction to QC concepts and to illustrate their relevance and potential impact on radiotherapy inverse planning. The core challenge in radiotherapy planning is the solution of large-scale optimization problems, balancing target coverage with sparing of organs-at- risk (OARs). Despite decades of algorithmic development, inverse planning still presents challenges. Current optimization algorithms, whether deterministic or stochastic, are typically designed to converge toward a single solution. However, the inverse planning problem is intrinsically degenerate: for any given case, multiple treatment plans may satisfy the optimization objectives equally well, each representing a different trade-off between competing goals. In this context, QC offers a fundamentally different computational framework that exploits key principles of quantum mechanics, such as superposition and entanglement, enabling efficient exploration of
Made with FlippingBook - Share PDF online