Table 4.

Performance metrics of DNN model in diagnosing malignant nodules of different sizes, evaluated on normal subset versus HT subset.

		HT Subset	Normal subset
Average size (SD)		0.975 (0.51)	1.25 (0.77)
<5 mm	AUC	0.915	0.895
	Accuracy	0.83	0.825
	Sensitivity	0.859	0.82
	Specificity	0.828	0.826
	Precision	0.327	0.651
5–10 mm	AUC	0.909	0.895
	Accuracy	0.82	0.846
	Sensitivity	0.902	0.868
	Specificity	0.794	0.822
	Precision	0.577	0.841
10–20 mm	AUC	0.883	0.907
	Accuracy	0.832	0.837
	Sensitivity	0.854	0.878
	Specificity	0.824	0.792
	Precision	0.652	0.827
>20 mm	AUC	0.871	0.845
	Accuracy	0.836	0.801
	Sensitivity	0.722	0.724
	Specificity	0.864	0.837
	Precision	0.594	0.688

AUC, Areas under the ROC curve. All metrics were the average of 10-folds.