Evaluation of the plan delivery accuracy of intensity-modulated radiation therapy by texture analysis using fluence maps.

Purpose: The aim of this study is to evaluate the plan delivery accuracy of intensity-modulated radiation therapy (IMRT) by using textural features calculated from fluence maps.

Methods: In total, 202 IMRT plans were generated with the Trilogy (C-series linac) and TrueBeam STx systems. Fluence maps were obtained for each IMRT plan. With the fluence map, six textural features were calculated, namely angular second moment (ASM), inverse difference moment (IDM), contrast, variance, correlation, and entropy (particular displacement distances, d = 1, 5, and 10). To evaluate performances of textural features in predicting IMRT delivery accuracy, Spearman's rank correlation coefficients (r) were calculated between textural feature values and measures of conventional methods.

Results: For Trilogy, correlation (d = 1) showed r values of -0.764 (p < 0.001) with gamma passing rates of 2%/1 mm and 0.863 (p < 0.001) with multi-leaf collimator (MLC) errors. For TrueBeam STx, variance (d = 10) showed r values of 0.691 (p < 0.001) with gamma passing rates of 1%/2 mm and -0.585 (p < 0.001) with MLC errors.

Conclusion: Correlation (d = 1) for Trilogy and variance (d = 10) for TrueBeam STx demonstrated the potential in predicting IMRT delivery accuracy with high correlations to conventional IMRT delivery accuracy measures.