Table 2. The classification performance of our proposed method for 14 diseases.

Disease	Accuracy	Precision	Sensitivity	F1 score
Atelectasis	0.722±0.004	0.613±0.007	0.733±0.008	0.669±0.004
Cardiomegaly	0.865±0.003	0.727±0.014	0.810±0.013	0.766±0.011
Effusion	0.796±0.010	0.695±0.009	0.835±0.016	0.758±0.010
Infiltration	0.593±0.006	0.551±0.009	0.787±0.007	0.648±0.007
Mass	0.858±0.008	0.471±0.010	0.694±0.014	0.561±0.011
Nodule	0.850±0.005	0.622±0.012	0.664±0.012	0.642±0.010
Pneumonia	0.798±0.013	0.271±0.014	0.490±0.009	0.349±0.008
Pneumothorax	0.867±0.012	0.638±0.007	0.766±0.009	0.696±0.005
Consolidation	0.581±0.009	0.234±0.019	0.776±0.010	0.360±0.012
Edema	0.933±0.010	0.229±0.018	0.467±0.010	0.307±0.012
Emphysema	0.926±0.004	0.312±0.012	0.960±0.007	0.471±0.008
Fibrosis	0.886±0.005	0.153±0.012	0.367±0.018	0.216±0.016
Pleural thickening	0.867±0.007	0.275±0.008	0.630±0.014	0.383±0.012
Hernia	0.849±0.010	0.119±0.001	0.970±0.001	0.212±0.001
Mean	0.814±0.003	0.422±0.010	0.714±0.008	0.503±0.006

The results of mean ± standard deviation are reported by randomly splitting the training and validation sets for ten times.