Table 3.

Evaluation of model performance using the testing set of patients who developed grade 4 radiotherapy-induced lymphopenia.

	Hybrid deep learning modela	Logistic regression	Logistic regression with elastic-net regularization	Random forest	Support vector machines
Accuracy	0.769	0.717	0.722	0.718	0.699
F1 score	0.695	0.616	0.589	0.647	0.663
Precision	0.740	0.681	0.621	0.651	0.604
Recall	0.656	0.563	0.575	0.644	0.736

^aUsing stacked bi-directional long-short term memory architecture for the dose-volume histogram input.