Operational Insights From the Longitudinal Analysis of a Linear Accelerator Machine Log.

Purpose This study aimed to perform a longitudinal analysis of linear accelerator (linac) technical faults reported with a cloud-based Machine Log system in use in a busy academic clinic and derive operational insights related to linac reliability, clinical utilization, and performance. Methods We queried the Machine Log system for the following parameters: linac type, number of reported technical faults, types of fault, number of faults where the linac was disabled, and estimated clinical downtime. The number of fractions treated and monitor units (MU) delivered were obtained from the record and verify system as metrics of linac utilization and to normalize the number of reported linac faults, facilitating inter-comparison. Two Varian TrueBeam C-arm linacs (Varian Medical Systems, Palo Alto, CA), one Varian 21iX C-arm linac (Varian Medical Systems, Palo Alto, CA), and one newly installed Varian Halcyon ring gantry linac (Varian Medical Systems, Palo Alto, CA) were evaluated. The linacs were studied over a 30-month period from September 2017 to March 2020.  Results Over 30 months, comprising 677 clinical days, 1234 faults were reported from all linacs, including 153 "linac down" events requiring rescheduling or cancellation of treatments. The TrueBeam linacs reported nearly twice as many imaging, multileaf collimator (MLC), and beam generation faults per fraction, and MU as the Halcyon. Halcyon experienced fewer beam generation/steering, accessory, and cooling-related faults than the other linacs but reported more computer and networking issues. Although it employs a relatively new MLC design compared to the C-arm linacs and delivers primarily intensity-modulated treatments, Halcyon reported fewer MLC faults than the other linacs. The 21iX linac had the fewest software-related faults but was subject to the most cooling-related faults, which we attributed to extensive use of this linac for treatment techniques with extended beam-on times. Conclusions A longitudinal analysis of a cloud-based Machine Log system yielded operational insights into the utilization, performance, and technical reliability of the linacs in use at our institution. Several trends in linac sub-system reliability were identified and could be attributed to either age, design, clinical use, or operational demands. The results of this analysis will be used as a basis for designing linac quality assurance schedules that reflect actual linac usage and observed sub-system reliability. Such a practice may contribute to a clinic workflow subject to fewer disruptions from linac faults, ultimately improving efficiency and patient safety.