LAUSR

Purpose The aim of this study was to propose a method of mean lung dose (MLD) prediction for III non-small cell lung cancer patients (NSCLC), based on their individual anatomy. The method was validated by comparison with results for volumetric modulated arc therapy (VMAT) plans, obtained with MCO (RayStation, v 5.1) and PRO (Eclispe, v13). Methods Dose distributions calculated for each patient in Eclispe for a set of single fields and the method based on linear equations were implemented in a standalone, homemade software (DosePredictor). The software predicts MLD. Prediction results were validated for a group of 21 patients with NSCLC treated in our clinic. Coplanar dynamic two full arcs VMAT plans were prepared with MCO and PRO for each patient. Prescribed dose (PD) of 58,80 Gy was delivered in 21 fractions. For all techniques, the same objectives were used: 95% of PD covering at least 98% of planning target volume, minimization of MLD. The Wilcoxon signed pairs rank test was used to compare predicted MLDs and the ones obtained by MCO and PRO. Correlation between DosePredictor and VMAT plans was examined using Pearson correlation coefficient. Results The average MLD was 13,7 Gy [8,5 Gy–20,8 Gy]; 14,0 Gy [8,5 Gy–2,6 Gy]; 14,8 Gy [8,7 Gy–24,3 Gy] for DosePredictor, PRO and MCO respectively. There was no significant difference between predicted MLD and the one calculated with PRO. For MCO the difference was significant ( α  = 0,01), but small (1,1 Gy ± 1,2 Gy). The Pearson correlation coefficient was 0,918 and 0,925 for MCO and PRO respectively. Conclusions The method allows to predict MLD for individual patient without necessity of plan preparation. The method can be used to define starting constraints for VMAT optimization or as a guidance for less skilled treatment planners.