Machine-learning-based prediction of the effectiveness of the delivered dose by exhale-gated radiotherapy for locally advanced lung cancer: The additional value of geometric over dosimetric parameters alone.

Abstract

This study aimed to assess interfraction stability of the delivered dose distribution by exhale-gated volumetric modulated arc therapy (VMAT) or intensity-modulated arc therapy (IMAT) for lung cancer and to determine dominant prognostic dosimetric and geometric factors. Clinical target volume (CTVPlan) from the planning CT was deformed to the exhale-gated daily CBCT scans to determine CTVi, treated by the respective dose fraction. The equivalent uniform dose of the CTVi was determined by the power law (gEUDi) and cell survival model (EUDiSF) as effectiveness measure for the delivered dose distribution. The following prognostic factors were analyzed: (I) minimum dose within the CTVi (Dmin_i), (II) Hausdorff distance (HDDi) between CTVi and CTVPlan, (III) doses and deformations at the point in CTVPlan at which the global minimum dose over all fractions per patient occurs (PDmin_global_i), and (IV) deformations at the point over all CTVi margins per patient with the largest Hausdorff distance (HDPworst). Prognostic value and generalizability of the prognostic factors were examined using cross-validated random forest or multilayer perceptron neural network (MLP) classifiers. Dose accumulation was performed using back deformation of the dose distribution from CTVi to CTVPlan. Altogether, 218 dose fractions (10 patients) were evaluated. There was a significant interpatient heterogeneity between the distributions of the normalized gEUDi values (p<0.0001, Kruskal-Wallis tests). Accumulated gEUD over all fractions per patient was 1.004-1.023 times of the prescribed dose. Accumulation led to tolerance of ~20% of fractions with gEUDi <93% of the prescribed dose. Normalized Dmin >60% was associated with predicted gEUD values above 95%. Dmin had the highest importance for predicting the gEUD over all analyzed prognostic parameters by out-of-bag loss reduction using the random forest procedure. Cross-validated random forest classifier based on Dmin as the sole input had the largest Pearson correlation coefficient (R=0.897) in comparison to classifiers using additional input variables. The neural network performed better than the random forest classifier, and the gEUD values predicted by the MLP classifier with Dmin as the sole input were correlated with the gEUD values characterized by R=0.933 (95% CI, 0.913-0.948). The performance of the full MLP model with all geometric input parameters was slightly better (R=0.952) than that based on Dmin (p=0.0034, Z-test). Accumulated dose distributions over the treatment series were robust against interfraction CTV deformations using exhale gating and online image guidance. Dmin was the most important parameter for gEUD prediction for a single fraction. All other parameters did not lead to a markedly improved generalizable prediction. Dosimetric information, especially location and value of Dmin within the CTV i , are vital information for image-guided radiation treatment.