Der Einfluss von Kollimatorlamellensequenzierung und Fehler in der Dosimetrie kleiner Felder auf die dosimetrische Genauigkeit von Behandlungsplänen in der Strahlentherapie

Abstract

In this work, the influence of errors in the dosimetry of small fields on dosimetric treatment plan quality was investigated for four different anatomical sites. Brain metastases, prostate including lymph nodes, prostate boost and head & neck were considered in the study. The measured dose distributions of the treatment plans were compared with calculated dose distributions generated by iPlan (Brainlab AG, Feldkirchen, Germany) (IMRT) and Monaco (Elekta, Stockholm, Sweden)(VMAT) using a gamma pass rate analysis (GPR). The calculated dose distributions were based on a clinical used beam model (BM) and a beam model with erroneous output factors (OF), designed to account for the influence of uncertainties in small field dosimetry. The measurements were made using radiochromic films. Additionally a three-dimensional GPR-analysis comparing the calculated doses of both beam models using 3D Slicer, a widely used software tool in radiation therapy (RT) was conducted. In order to map small field dosimetry errors and the resulting GPR to the specific geometry of a treatment plan, DICOM data containing treatment plan specifications were analyzed and the aperture distribution for each plan was calculated. Out of this, a model based on the difference in OFs between the clinical BM and the modified BM was designed and implemented into Matlab. Correlations were analyzed between the error factors generated by this model and the GPR result based on the clinically used beam model. Furthermore, a correlation analysis of the GPR based on comparison of measurement with clinical BM and modified BM was conducted. The overall GPR using a criterion of 3 % / 3 mm conducted with 3D Slicer was well above 95 %. The GPRs based on comparisons of calculated and measured dose of both BMs using the same criterion, was mostly > 90 %, except 5 outliers out of 46 considered treatment plans. There was no correlation found for any treatment group evaluated between the error factors generated by the above mentioned model and the GPR results based on the clinical BM. However, there was a significant correlation found for 2 treatment groups between the GPR based on the modified BM and the clinical BM, both showing better results for the modified BM. Finally, the general dependencies and restrictions are presented and an additional error model based on Clarkson integration is outlined. In this work it was found, that small field errors did not translate into unacceptable GPR. It was also found, that the BM modification lead to sensitive areas in the region of steep dose gradients, but so far a reason for this could not be investigated. Furthermore, the assessment of error factors is not trivial and the proposed model to generate error factors for an arbitrary leaf constellation probably needs further improvement. The concept of GPR-analysis was also discussed in detail and it was concluded, that the method has its difficulties in terms of comparability.