Utilising elekta LINAC and MLC controller log files for phantom-less patient specific IMRT QA

Abstract

Introduction: Intensity-modulated radiation therapy (IMRT) is the state of the art treatment technique in external beam radiotherapy. Currently, quality assurance (QA) procedures rely mostly on phantom-based experimental measurements. However, electronic log files, which originate from the linear accelerator (LINAC) or multi leaf collimator (MLC) controller, offer promising novel options for QA. Using this methodology, a variety of machine parameters can be recorded during the treatment. The aim of this MSc thesis was to explore and validate this log file-based QA. Materials and Methods: Two different models of LINACs, Versa HD and Synergy (Elekta AB, Stockholm, Sweden), were used to perform three test plans on a weekly basis: a 4-field-box, a head & neck volumetric-modulated arc therapy (VMAT) as well as a static picket fence (PF) test. The study duration was six months. XML type log files (MLC leaf positions, electric gun current etc.) were acquired during every irradiation using the option service graphing of the machine¿s service mode. An in-house software was applied to evaluate deviations ofcthe leaf positions from the test plan (as defined in the DICOM file). A mean leaf positioning error was calculated as the average of a mean measurement bank error over all measurements. The integrated electronic portal imaging device (EPID) and a detector array (Delta4, ScandiDos AB, Uppsala, Sweden) provided reference data. The sensitivity of all measurement methods was analysed by means of implementing leaf errors in the static and also dynamic PF test plan. Results: For the static plans (PF test, 4-field-box), Versa HD log files showed a lower mean leaf positioning error (0.010 mm X1/ -0.002 mm X2, 0.002 mm X1/ -0.042 mm X2) compared to the Synergy (-0.301 mm/ 0.309 mm, -0.296 mm/ 0.291 mm). The EPID obtained higher values for the standard deviation than the log files, which could be attributed to the finite resolution of the EPID panel (pixel size of 0.4 mm). Interestingly, no agreement in the time development of the mean measurement bank error could be observed between log file and EPID data (Pearson correlation coefficient around 0.3; even smaller values were obtained between log files and Delta4 VMAT measurements: 0.06). Implemented leaf errors of 0.5, 1 and 2 mm could be detected by the EPID as well as the log files. A rather small but significant influence of gravity was noticed by the log file analysis on the Versa HD. The gun current declined by 0.08 mA on Versa HD and 0.22 mA on Elekta Synergy, respectively. Discussion/Conclusion: Log files appear to be able to provide a variety of useful data for every treatment delivery performed, e.g. influence of gravity, decrease in gun current. Although the log files did not reflect the time behaviour monitored by the EPID and the Delta4 phantom on a long-time basis, it was possible to detect artificially introduced leaf errors. The simultaneous use of an independent system monitoring the leaf calibration is indicated. Once generally available on Elekta machines, log files might have the potential to act as an additional QA procedure.