Insights into Imaging

Table 4 Predictive performance of models 1, 2, and 3

From: Development and validation of MRI-based model for the preoperative prediction of macrotrabecular hepatocellular carcinoma subtype

Model	Training cohort						Validation cohort
Model	AUC (95% CI)	Sen	Spec	Accu	Cutoff	p value	AUC (95% CI)	Sens	Spec	Accu	Cutoff	p value
Model 1	0.773 (0.696–0.838)	0.521	0.910	0.722	0.438	Model 1 vs Model 2:0.548	0.801 (0.681–0.891)	0.619	0.904	0.702	0.524	Model 1 versus Model 2:0.089
Model 2	0.747 (0.689–0.806)	0.938	0.515	0.727	0.484		0.718 (0.618–0.810)	0.857	0.500	0.679	0.357
AP	0.706 (0.642–0.766)	0.526	0.742	0.634	0.351	Model 1 vs Model 3: 0.001	0.589 (0.486–0.699)	0.571	0.595	0.583	0.381	Model 1 versus Model 3: 0.321
PVP	0.662 (0.600–0.731)	0.557	0.629	0.593	0.268	Model 2 vs Model 3:0.001	0.638 (0.538–0.738)	0.643	0.571	0.607	0.381	Model 2 versus Model 3:0.031
T2W	0.673 (0.610–0.736)	0.722	0.608	0.665	0.340		0.580 (0.470–0.681)	0.548	0.524	0.536	0.357
Model 3	0.889 (0.851–0.926)	0.866	0.784	0.825	0.648		0.866 (0.801–0.928)	0.881	0.714	0.798	0.548

AP arterial phase, PVP portal venous phase, T2W T2-weighted image, AUC area under the curve, Sen sensitivity, Spec specificity, Accu accuracy, CI confidence interval

Back to article page