Presenter(s)
J. Charters1, and X. Qi2; 1UCLA, Department of Radiation Oncology, Los Angeles, CA, 2Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA
Purpose/Objective(s): Online adaptive MR-guided radiotherapy (MRgRT) involves complex, multi-stage workflows requiring coordinated multidisciplinary effort. While overall adaptive treatment durations have been reported, the relative contribution and variability of individual workflow components across large clinical cohorts remain poorly characterized. This study quantifies event-specific time allocation and interruption burden during clinical adaptive MRgRT using an automated log file analysis approach.
Materials/Methods: Machine log files from a clinical 0.35 T MR-linac were collected between March and October 2025. Log files were parsed to identify time-stamped workflow events corresponding to patient positioning, contouring, manual planning, plan selection, adaptation approval, and treatment initiation. Event durations were aggregated across adaptive sessions performed during routine clinical operation. For each session, total adaptive duration and the percentage of time spent in each workflow stage were calculated. Variability was assessed using descriptive statistics and statistical process control (SPC) metrics. Treatment interruptions, including pauses and early termination events, were quantified by frequency and associated time impact.
Results: Across N = 82 adaptive sessions, median total adaptive workflow duration, from session initiation to treatment start, was 41.25 minutes (IQR 31.43 – 47.89 minutes). Online re-contouring (median 13.27 minutes, 35.31%) accounted for the largest proportion of adaptive time, followed by patient positioning and imaging approval (median 8.48 minutes, 27.12%). Manual planning and plan selection together comprised 10.01% of total adaptive duration. The greatest relative variability was observed during cine imaging and target tracking preparation after plan approval (CV = 160.68%). Overall, 12.20% of sessions exceeded 60 minutes, and 1.22% exceeded the upper control limit (3 standard deviations above the mean). Treatment interruptions occurred in 18.29% of sessions, and 25.61% of treatments ended early.
Conclusion: Automated log file analysis enables scalable, objective quantification of event-specific adaptive workflow burden in clinical MRgRT. Adaptive time is primarily driven by patient positioning and contouring, with substantial variability and measurable interruption impact. When integrated with SPC methods, this approach enables monitoring of workflow stability, identification of deviations, and supports data-driven quality improvement. These findings provide actionable metrics to optimize efficiency and resource allocation in adaptive radiotherapy practice.