ESTRO 36 Abstract Book

S423 ESTRO 36 _______________________________________________________________________________________________

PO-0795 Comparison of Service graph log and Dynamic linac log of Elekta Linacs for patient QA. M. Kowatsch 1 , M. Meinschad 1 , G. Leitold 1 , P. Szeverinski 1 , T. Künzler 1 1 LKH Feldkirch, Institut of Medical Physics, Feldkirch, Austria Purpose or Objective The complexity of intensity modulated radiation therapies (IMAT, IMRT) requires patient specific pretreatment verification of calculated dose distributions which is time consuming. Elekta linacs provide 2 different log files. One is the Service graph (SG) with a resolution of 4 Hz and is directly accessible through the service mode on the linac. The second one is the Dynamic linac log (DLL) with a resolution of 25 Hz. The aim of this study is to compare both types of log files for dose recalculation with Monte Carlo and beam statistics for an Elekta Synergy linac with Agility MLC (Elekta, Crawley). Material and Methods To compare the log files 2 head & neck, a mamma left side, an abdomen with simultaneous integrated boost, a thoracic spine with 3 dose levels and 1 brain case were chosen. Different parameters like leaf travel (LT), the sum of travel of all leaves between the open jaws, leaf speed (LS), leaf position (LP) and modulation complexity score (MCS) (Masi, Med. Phys. 40, 071718, 2013) were compared between the SG and the DLL. The DICOM RT file was used as reference for comparing LT and MCS. Furthermore log files were converted with an in-house Matlab script to .tel files to recalculate the irradiated plans with Monaco 5.0 TPS (Elekta, Crawley). For recalculation a grid size of 3mm and an uncertainty of 1% per control point were used resulting in a final uncertainty of roughly 0.1%. Isodose and DVH comparison were performed to evaluate equality of recalculated and originally calculated plans. Results The difference for leaf travel between SG and DLL to the Dicom-RT file was between -9.5% to 2.7% and -0.4% to 6.2%, respectively and between SG and DLL from -2.8 to - 11.3%. The differences of the MCI between the two log files was -0.4% to 0.3% and up to 20% compared to the DICOM file (see Table 1). The difference of 20% for plan 6 originates from the definition of LT. In this case, 2 beams with 2 arcs were evaluated. For SG and DLL all beams were evaluated as a single beam, the Dicom RT files were evaluated beam-by-beam. The maximum LT for a particular leaf between 2 control points (CP) showed big discrepancies and was in one case 20.1 mm for the SG and 32.6 mm for the DLL. The differences originate from writing errors between CPs in the SG and these errors are still inexplicable.Random dose errors in DVH up to +-0.5 Gy can be seen by recalculation of both log files for the entire plan. For linac parameter statistics (LT, LS, LP) the SG cannot be used because of random writing errors.

e.g. in calculation, positioning and movement, spatial precision and absolute dose application. We present a test that was introduced into the clinical workflow and evaluated its sensitivity to those errors. Material and Methods Prior to the irradiation, a custom-built phantom insert for the ArcCHECK (Sun Nuclear, USA) allowed for automatic registration of the cone beam CT to reference data. A 12- field plan including gantry and table rotations targeting a spherical volume of approx. 2 cm diameter was measured weekly using a Synergy accelerator with an Agility MLC (Elekta, Sweden). Signals were obtained from all diodes along the cylinder surface of the ArcCHECK and additional dose was measured with an ionization chamber in the phantom center. For each measurement the plan was compared to the calculation of the treatment planning system via gamma evaluation and every diode reading was compared to the averaged diode readings from previous weeks. Additionally, errors were induced to test the sensitivity for phantom malposition, machine geometry problems and MLC positional inaccuracies. Results Due to the phantom set up according to the cone beam CT registration, the measurements were very reproducible without any observable user-to-user differences. The typical dose map for the diode cylinder is shown in fig. 1. For all diodes, mean values with small standard deviations were obtained from many consecutive measurements. Any diode deviation observed for the correct application of the test plan never exceeded three standard deviations, while much larger discrepancies could be detected for all induced errors (example: fig. 2).

Conclusion We developed a fast end-to-end test for stereotactic radiation therapy with the ArcCHECK phantom which minimizes user influences for high reproducibility and was easily included into clinical routine. It compares the dose distribution on a helical diode array and a cumulative central dose with the doses from the treatment planning system. By additionally comparing each of the over 1300 diode values to a corresponding average dose derived from previous measurements, the method simultaneously serves as a constancy test of all involved components and is able to reliably detect a vast variety of even very small errors.

Conclusion Both file types are accurate for dose recalculation. The 4 Hz resolution and writing errors of the Servicegraph log are limiting a robust statistical analysis of linac parameters. Dynamic linac logs allow for dose recalculation and for a more detailed statistical analysis of the linac. Both types of log files can be taken for patient QA to decrease the workload of measurements and for