S1601
Physics - Autosegmentation
ESTRO 2026
Over individual follow-up, 6 patients experienced overall decline in TMT, and 7 showed an overall increase (Figure 2A). A significant relationship between TMT and age was observed for both sexes (p<0.05, Figure 2B).
Digital Poster Highlight 4310
Accelerating treatment pathways with AI auto- contouring: Reducing time from CT to treatment through automation and dynamic patient scheduling Michael Roche, Morgan Healy, Patricia Coen, Michael Moran, Orla McKivitt, R B Ezhilalan, Eoin McGrath, Martin Higgins CUH/UCC Cancer Centre, Cork University Hospital, Cork, Ireland Purpose/Objective: Radiotherapy departments are continuously adopting technological solutions to improve workflow efficiency for patient treatments. Improved efficiencies in real world contouring tasks using AI-based contouring systems have been reported, with shorter active contouring task and workflow times.1 However, the assumption that task level efficiencies automatically translate into faster overall patient pre-treatment pathways has been challenged.2 This single institution multidisciplinary study aims to evaluate how AI based contouring and dynamic patient scheduling can be used to accelerate the pathway from CT to treatment. Material/Methods: Workflow data was retrospectively evaluated over three years (2023-2025) both before and after the introduction of the MVision Contour+TM auto- contouring software. Data evaluated included radiotherapy schedules for 3722 patients. Radiotherapy task completion times were recorded from the Quality Check Lists (QCLs) available from the oncology information system. The QCLs record a timestamp for each step in the radiotherapy workflow, from CT to contour completion, planning and first treatment. The contouring workflow time and the overall CT to treatment interval were evaluated pre and post implementation of the auto-contouring software. Patient start dates were scheduled post treatment planning to optimize the CT to treatment workflow time. Results: The mean contouring time for patients went from 5.2 ± 5.1 calendar days (95% CI 4.9-5.5) in 2023 to 3.9 ± 3.8 days (95% CI 3.7-4.1) and 3.7 ± 3.7 days (95% CI 3.4- 3.9) in 2024 and 2025 respectively, post implementation of MVision Contour+TM. The mean CT to treatment interval was 17.1 ± 8.1 calendar days (95% CI 16.6-17.6) in 2023 compared to 15.6 ± 6.8 days (95% CI 15.3-16) and 14.8 ± 6.7 days (95% CI 14.4-15.2) in 2024 and 2025 respectively. In 2023, 71% of patients started treatment within 21 calendar days of CT, compared to 83% and 88% in 2024 and 2025. The percent of contours completed weekdays from 6pm to 8am and weekend days, decreased from 42% in 2023 to 18% across 2024/2025.
Conclusion: Temporalis muscle thickness from routine MRI is a promising biomarker for monitoring sarcopenia in childhood and TYA brain cancer survivors. Our findings suggest it is feasible to measure TMT using existing auto-segmentation tools. Such measurements could help identify patients who may benefit from
targeted follow-up care, such as, exercise programmes, hormone supplementation, or
nutritional support. Further optimisation is needed to improve TMT estimation in complex cases. While our current approach identified obvious failures, more detailed error analysis will be explored in future work involving radiology and oncology expertise. References: [1] L A Gilligan et al., Paediatric Radiology, 2019, https://doi.org/10.1007/s00247-019-04562-7.[2] A Zapaishchykova et al., Nature Communications, 2023, https://doi.org/10.1038/s41467-023-42501-1.[3]M Mandzak et al., arXiv, 2025,
https://doi.org/10.48550/arXiv.2506.05660 Keywords: sarcopenia, paediatric, MRI
Made with FlippingBook - Share PDF online