Linac performance benchmarking: Feasibility of a web‐based tool for comparison of local machine performance check (MPC) results with the global community
Lauren May, Michael P. Barnes, Michael Pearson, Benjamin Zwan, José Antonio Baeza‐Ortega, Benjamin Cooper, Mohammad Hussein, Ajeet Mishra, Shiaw Juen (Eugene) Tan, Jackson Zifodya, Peter B. Greer, Catharine H. Clark, Joerg LehmannAbstract
Background
Varian machine performance check (MPC) is a manufacturer integrated quality control system included as standard for TrueBeam and Halcyon linear accelerators (linacs) to assess performance using a broad range of geometric and beam tests. MPC is a highly standardized method of performance testing, is generally free of inter‐user variability and is quick and easy to perform allowing daily acquisitions. Therefore, previous studies have highlighted the potential of comparative analysis of MPC data in investigating differences in performance of individual linacs and identifying outliers. Statistical process control (SPC) is an established approach commonly used to assess stability and behavior of sequential data and identify standard operating ranges which has previously been applied to radiotherapy quality assurance data and MPC data.
Purpose
This study aimed to investigate the feasibility of a web‐based platform using SPC and aggregated MPC data to allow users to compare their machine performance against that of a cohort of others in the radiotherapy community.
Methods
A previously developed web‐based platform for radiotherapy data sharing was expanded to include the MPC data collection and comparison tool. Pilot centers exported data from their MPC applications and uploaded it through center‐specific portals. Within the application, SPC analysis was used to calculate upper control limits (UCL), lower control limits (LCL) and central lines (CL) for the data sets of each individual linac as well as cohort averages. Interactive plots of longitudinal data for individual machines were displayed for users and overlaid with SPC limits for the specific machine and the cohort averages. Additional analyses were conducted externally from the app where the average and range of SPC control limits for the cohort were compared against MPC tolerances.
Results
MPC data for 57 648 measurements taken between April 2016 and May 2026 were collected across seven centers from 21 TrueBeam linacs (15 with Millennium multi‐leaf collimator (MLC), six with HD120 MLCs) and one Halcyon. Variation in performance for specific tests was identified for individual machines compared against the cohort as well as for individual MLC leaves for specific machines. Across the cohort of TrueBeam linacs, average control ranges for all geometric tests were within MPC thresholds. However, the range of values for UCL and LCL exceeded tolerances for relative gantry angle. Machines installed before 2021 were found to have slightly larger variability in mean MLC offset with an average control range width of 0.20 mm and 0.18 mm for bank A and B, respectively, compared to machines installed after 2021 of 0.10 mm and 0.13 mm. Analysis of data for individual linacs segmented per year showed a change in mean MLC offset occurring around 2021.
Conclusions
The usefulness and feasibility of a tool to compare MPC data against the radiotherapy community was demonstrated. Inter‐center comparison of data allowed trends to be identified which would otherwise not be observable for individual centers. Further development and distribution of the MPC data sharing and comparison tool may provide improved confidence and understanding of machine performance and assist low‐resourced clinics in delivering safe and accurate radiotherapy treatments.