S2907
RTT- RTT operational practice and workflow innovations
ESTRO 2026
reduced hallucination risk and improved the reliability of LLM-generated responses. This approach demonstrates a feasible pathway for integrating AI- based clinical assistants into routine radiation therapy practice, supporting standardisation, training, and just- in-time information access. Keywords: Artificial Intelligence, Chatbot, LLM
domain, link the relevant work-instruction documents, and to utilise guardrails to minimise misinterpretation of common abbreviations and reduce hallucinations. Reference SOPs were reformatted into markdown with clear headings to facilitate instruction-based citation.We created sub-domain specific chatbots dedicated to CT simulation and treatment workflows (photon and proton), thereby improving relevance by narrowing the knowledge scope for each assistant. The chatbots underwent iterative testing using queries of increasing complexity; based on observed responses, both prompt instructions and reference documents were refined to enhance accuracy and consistency of domain specific answers. Results: Customised Pair Assistant chatbots were able to address CT simulation and treatment-related procedural queries with high precision. The assistants demonstrated the capability to request additional contextual details before providing an answer. An infographic summarising the development and deployment workflow is presented in Figure 1a, and utilisation metrics are shown in Figure 1b.
Proffered Paper 1369
Quantifying RTT workload in Proton Research: Insights from Evaluative Commissioning Studies. Riya Patel 1 , Amanda Webster 1 , Laura Allington 1 , Yen Ching Chang 1 , Sally Falk 2 , Faye Hellewell 2 , Syed Moinuddin 1 , Ed Smith 2 , Danielle Fairweather 1 1 University College London Hospitals NHS Foundation Trust, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 2 The Christie NHS Foundation Trust, The Christie NHS Foundation Trust, Manchester, United Kingdom Purpose/Objective: The Evaluative Commissioning in Protons (ECIP) programme is a national NHS England initiativeestablishedto assess new indications for PBT whereRCTs are not feasible. RTTs are integral to the coordination and delivery of these complex studies, extending their contribution beyond traditional treatment delivery. This evaluation examines RTT workload associated with ECIP study activities and highlights the evolving scope of RTT-led research delivery. Material/Methods: A prospective time audit was undertaken at one of the two PBT centres, to quantify RTT workload associated with ECIP study activities. RTTs recorded the duration of all study-related tasks per participant across screening, enrolment, follow-up stages. Tasks were grouped into predefined domains: patient-facing activities, data entry, documentation, regulatory, and coordination. Each record captured staff role, task type, visit type, and duration. Qualitative notes provided context for RTT specific requirements. Data was analysed descriptively to determine mean time per patient and task category, and to identify process bottlenecks and staffing requirements. Results: Three studies received researchethicsapproval between November 2024 and April 2025. The mean time from site set-up to opening was 278 days ( ≈ 9 months), with the first two studies opening in September and October 2025. Projected recruitment across these studies for a single centre is 115 patients
Potential hallucinations were mitigated through explicit behavioural instructions, such as: “Be conservative: if any part of the query is uncertain or not explicitly stated, do not infer or guess,” and “Only use information explicitly contained in the reference document.” The chatbot was also configured to prompt the user for more information when inputs were ambiguous. For queries outside its knowledge base, it explicitly acknowledged its limitations and declines to provide advice. Conclusion: We successfully developed dedicated chatbots that provide rapid, context-aware guidance for CT simulation and photon/proton radiotherapy workflows. By combining domain-specific prompt engineering with reference-based guardrails, we
Made with FlippingBook - Share PDF online