Session Item

The aim of this work is to find a simple and effective mathematical method to predict the output variations of TrueBeam (Varian Medical Systems, Palo Alto, CA, USA) linear accelerators (Linacs), using the daily output measurements performed with Machine Performance Check (MPC) tool.

Three TrueBeam Linacs version 2.7, all placed in Humanitas ICC radiotherapy department, were equipped with the aS1200 portal imager. The MPC application is an automated tool, integrated into all three Linacs, allowing to perform a check of beam output, uniformity and various geometric parameters, utilizing a dedicated phantom and the imaging system.

All daily output data, measured with MPC from April 2020 to October 2021, were exported and plotted for all energies to visualize the trend. The first step was to identify the dates of absolute dose calibration, performed in accordance with the IAEA TRS-398 protocol. Then, in order to generalize the results, the measurement period was divided into five sub-periods, common to the three Linacs, taking into account the calibration dates. Finally, a trend line was applied to the data of the series corresponding to each sub-period. Moreover, daily data were compared with monthly output measurements performed by means of ionization chambers and RW3 phantom.

The output of all the TrueBeam Linacs under examination, measured with both MPC and ionization chambers, showed an increase for all photon and electron energies (on average +4% per year), in agreement with the literature findings. Output data for the 6 MV energy are representative of all the energies and all the Linacs. Output calibrations and subsequent MPC re-baselines resulted in sudden drops of output towards baseline.

The regression line fitted the data quite well and represented the average Linac output. The angular coefficient of each line characterized the rate of output increase in the time interval considered. We found that the difference in output increase for all three Linacs and for each sub-period was not statistically significant (p>0.3), so it was possible to generalize using a universal angular coefficient that indicated the daily amount of output drift (on average 0.016%). Applying the linear mathematical model to our daily output data evaluated with MPC, the trends of Figure 1 were obtained. The overall output increase found with the model is 0.34% per month, comparable with the literature data.

In this study, linear regression analysis was used to predict the TrueBeam output drift over time and a unique factor identifying the rate of increase was found. Therefore, MPC with daily output measurements is a reliable tool in predicting TrueBeam output behavior and helping physicists to find the right timing for preventive beams calibration throughout the year. Further investigation is needed to eventually include in the analysis other factors affecting the output such as seasonal fluctuations and environmental factors.