Deep learning approach for early prediction of COVID-19 mortality using chest X-ray and electronic health records

BMC Bioinformatics

Table 2 Performance of each model, including the ensemble model

Method	AUROC	Accuracy	Precision	Recall	F1 score
Chest X-ray
Inception-ResNet-V2	0.6166	0.76	0.50	0.50	0.46
EfficientNet B2	0.6769	0.78	0.65	0.55	0.55
EfficientNet B1	0.7063	0.77	0.64	0.57	0.57
EHR data
XGBoost	0.8352	0.85	0.81	0.70	0.73
RF	0.7980	0.84	0.82	0.66	0.70
DL (MLP)	0.8109	0.84	0.79	0.68	0.71
Ensemble model^a	0.8698	0.84	0.86	0.66	0.69
Optimized ensemble model^b	0.8698	0.86	0.81	0.74	0.77

AUROC area under the receiver operating characteristic curve, EHR electronic health record, RF random forest, DL deep learning, MLP multi-layer perceptron
^aEnsemble model: Ensemble of EfficientNet B1, XGBoost, RF and DL (MLP)
^bEnsemble model was optimized by F1 score (cutoff-value adjustment)

ISSN: 1471-2105