Log File-Based Automated Parsing and Statistical Process Control of Adaptive MR-Guided Radiotherapy Workflows
Abstract
Purpose
MR-guided radiotherapy, especially adaptive therapy, involves complex, multi-stage clinical workflows that are challenging to quantify using conventional manual metrics. In-depth, quantitative, understanding of the clinical workflow can provide crucial insights to identify bottlenecks and improve operational efficiency. This work developed a machine log file-based automated tool to automatically derive clinically meaningful workflow timelines during the entire treatment process and to assess workflow quantitatively using statistical process control (SPC) methods.
Methods
Machine log files from a clinical 0.35 T MR-linac (March to October 2025) were extracted and parsed to identify time-stamped events corresponding to imaging, patient positioning, adaptive planning, and treatment delivery. These events were mapped to discrete workflow intervals, including MRI simulation duration, positioning and setup time, adaptive planning duration, geometry approval time, plan review time, beam-on time, and treatment interruptions. Time intervals were aggregated across treatment sessions and analyzed using control charts and descriptive statistics to quantify variability and identify outliers. Data acquired from machine commissioning and QA were excluded to reflect routine clinical operation.
Results
Automated parsing and log-derived workflow intervals enabled objective quantification of multiple components of the adaptive radiotherapy process without manual data collection. Adaptive planning-related intervals exhibited the greatest variability relative to imaging and delivery components. Control charts revealed stable operating ranges for several workflow stages while highlighting sessions exceeding expected duration thresholds. Across all analyzed MR simulations (N = 470), mean simulation duration was 19.31 minutes, with 8 (1.70%) cases exceeding 30 minutes. Across all analyzed adaptive sessions (N = 378), mean adaptive planning duration was 33.42 minutes, with 16 (4.23%) cases exceeding 1 hour.
Conclusion
Automated log file analysis enables scalable, objective characterization of MRgRT workflows. Together with SPC, the tool enables medical physicists to monitor workflow stability, identify deviations, and support data-driven quality assurance and continuous process improvement in MR-guided radiotherapy.