S2398
Physics - Quality assurance and auditing
ESTRO 2026
Peter W Fick, Jan Kubica, Wouter van Elmpt, Richard Canters, Dominique Reijtenbagh, Jonathan Martens, Marta Bogowicz Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, Netherlands Purpose/Objective: Automation and artificial intelligence are key drivers of efficiency in modern radiotherapy, enabling more patients to be treated while maintaining quality. Successful adoption requires accurate representation of clinical practice in script logic, and therefore continuous communication between radiation therapists (RTTs) and the software team. Since end - users often under-report software errors, log mining offers a novel solution. This work demonstrates the use of log mining tools to detect and analyze anomalies in automated plan checks. Material/Methods: The in-house developed Plan Check script automates 30 plan checks, covering nine clinical treatment indications. It has been in clinical use since December 2024 in a high - throughput department with around 100 planning RTTs across four clinical teams. The script can be run at any step of the planning process, displaying Pass/Fail/Not applicable/Warning/Error per protocol - defined item. Log files produced by the script are ingested into Splunk Enterprise (v10.0.1, Cisco), a popular log mining tool. Splunk enables continuous monitoring, querying, and dashboard creation. The Plan Check dashboard visualizes daily runs, status distribution, and protocol - specific usage, highlighting checks with high fail/error rates. Results from a one month period were collected and used for structured root - cause analysis for the check with the highest failure rate. Results: Logging of 461 Plan Check runs was analyzed. The HighResolutionCheck, which evaluates whether high - resolution contouring was applied to predefined anatomical structures, showed the highest failure incidence. Out of 2,508 entries (multiple anatomical structures evaluated per run), 119 failures were recorded.
Conclusion: Technological advancements and national standardisation efforts have markedly reduced OAR doses and the predicted dysphagia risk after radiotherapy of DAHANCA patients. These results demonstrate the substantial nationwide impact of guideline implementation, plan audits, and continuous quality improvement over more than a decade of modern radiotherapy practice. References: 1. Nielsen CP et al. Consistency in contouring of organs at risk by artificial intelligence vs oncologists in HNC patients. Acta Oncol. 20232. Konrad M et al. Automatic outlier detection in organ at risk and target delineation. Zenodo: 170761233. Van Den Bosch et al. Comprehensive toxicity risk profiling in radiation therapy for head and neck cancer: A new concept for individually optimised treatment. Radiother Oncol. 20214. Zukauskaite R et al. Comparison of 3-year local control using DAHANCA radiotherapy guidelines before and after implementation of five mm geometrical GTV to high-dose CTV margin. Radiother Oncol. 2024 Keywords: Time-trends, dysphagia
.Root - cause analysis required ~2 hours: one hour for data extraction via Splunk queries and one hour for interpretation. Causes were categorized into two domains: clinical practice (change in practice,
Digital Poster Highlight 4433 Monitoring anomalies in an automated process: a log mining approach
Made with FlippingBook - Share PDF online