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.

Purpose: 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.

Methods: Clinical target volume (CTV) from the planning CT was deformed to the exhale-gated daily CBCT scans to determine CTV, treated by the respective dose fraction. The equivalent uniform dose of the CTV was determined by the power law (EUD) and cell survival model (EUD) as effectiveness measure for the delivered dose distribution. The following prognostic factors were analyzed: (I) minimum dose within the CTV (D), (II) Hausdorff distance (HDD) between CTV and CTV, (III) doses and deformations at the point in CTV at which the global minimum dose over all fractions per patient occurs (PD), and (IV) deformations at the point over all CTV margins per patient with the largest Hausdorff distance (HDPw). 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 CTV to CTV.

Results: Altogether, 218 dose fractions (10 patients) were evaluated. There was a significant interpatient heterogeneity between the distributions of the normalized EUD values (<0.0001, Kruskal-Wallis tests). Accumulated EUD over all fractions per patient was 1.004-1.023 times of the prescribed dose. Accumulation led to tolerance of ~20% of fractions with EUD 93% of the prescribed dose. Normalized D >60% was associated with predicted EUD values above 95%. D had the highest importance for predicting the EUD over all analyzed prognostic parameters by out-of-bag loss reduction using the random forest procedure. Cross-validated random forest classifier based on D 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 EUD values predicted by the MLP classifier with D as the sole input were correlated with the EUD 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 D (=0.0034, Z-test).

Conclusion: Accumulated dose distributions over the treatment series were robust against interfraction CTV deformations using exhale gating and online image guidance. D was the most important parameter for EUD prediction for a single fraction. All other parameters did not lead to a markedly improved generalizable prediction. Dosimetric information, especially location and value of D within the CTV , are vital information for image-guided radiation treatment.