Prediction of VMAT delivery accuracy with textural features calculated from fluence maps

Abstract

Abstract Background Comprehensively textural feature performance test from volumetric modulated arc therapy (VMAT) fluences to predict plan delivery accuracy. Methods A total of 240 VMAT plans for various treatment sites were analyzed, with Trilogy and TrueBeam STx systems. Fluence maps superposed fluences at each control point per plan. The textural features were the angular second moment (ASM), inverse difference moment (IDM), contrast, variance, correlation, and entropy, calculated from fluence maps using three displacement distances. Correlation analysis of textural feature performance as predictors of VMAT delivery accuracy used global gamma passing rates with MapCHECK2 and ArcCHECK dosimeters, and mechanical delivery errors calculated from machine log files. Results Spearman’s rank correlation coefficients (r) of the ASM (d = 10) to the gamma passing rates with 1%/2 mm using the MapCHECK2 were 0.358 and 0.519, respectively (p <  0.001). For the ArcCHECK, they were 0.273 (p = 0.001) and 0.259 (p = 0.009), respectively. The r-values of the ASM (d = 10) to the Trilogy and TrueBeam STx MLC errors were − 0.843 and − 0.859, respectively (p <  0.001), and those to the MU delivery errors were − 0.482 and − 0.589, respectively (p <  0.001). The ASM (d = 10) showed better performance in predicting VMAT delivery accuracy. Conclusions The ASM (d = 10) calculated from VMAT plan fluence maps were strongly correlated with global gamma passing rates and MLC delivery errors, and can predict VMAT delivery accuracy.