Table 2.

Predictive performance of different DL signatures in training and validation sets

	Datasets	Phase	Sensitivity (%)	Specificity (%)	Accuracy (%)	AUC (95%CI)	p value
DL signature	Training set	AP	76.4	84.3	82.1	0.882 (0.823–0.941)	0.048*
		PVP	67.3	85.0	80.0	0.822 (0.753–0.891)	0.001*
		HBP	78.2	90.7	87.2	0.909 (0.860–0.958)	0.115
		mp-MR	76.4	93.6	88.7	0.929 (0.829–0.966)	Ref
	Validation set	AP	73.9	79.1	77.8	0.826 (0.755–0.897)	0.019*
		PVP	73.9	86.6	83.3	0.854 (0.780–0.928)	0.259
		HBP	78.3	77.6	77.8	0.888 (0.833–0.943)	0.743
		mp-MR	78.3	89.6	86.7	0.894 (0.811–0.977)	Ref

AUC, area under the curve; CI, confidence interval; AP, arterial phase; DL, deep learning; PVP, portal venous phase; HBP, hepatobiliary phase; mp-MR, multiple sequences magnetic resonance. p^* < 0.05 indicates a significant difference