Table 5.

Performance of the best MRI- and CT-based machine learning classification model

Data	ICG-R15	Cohort	AUC (95%CI)	Accuracy	Sensitivity	Specificity	model
MRI	ICG-R15 ≤ 10% vs. ICG-R15>10%	Training	0.996(0.989–1.000)	0.987	0.980	1.000	XGBoost
		Test	0.899(0.784–1.000)	0.853	0.875	0.833	XGBoost
	ICG-R15 ≤ 20% vs. ICG-R15>20%	Training	0.995(0.986–1.000)	0.962	0.929	0.980	Random Forest
		Test	0.979(0.941–1.000)	0.882	1.000	0.857	Random Forest
	ICG-R15 ≤ 30% vs. ICG-R15>30%	Training	0.997(0.991–1.000)	0.962	1.000	0.951	XGBoost
		Test	0.961(0.890–1.000)	0.941	1.000	0.968	XGBoost
CT	ICG-R15 ≤ 10% vs. ICG-R15>10%	Training	0.998(0.995–1.000)	0.970	0.957	1.000	XGBoost
		Test	0.822(0.700–0.944)	0.842	0.917	0.714	XGBoost
	ICG-R15 ≤ 20% vs. ICG-R15>20%	Training	0.866(0.781–0.951)	0.842	0.872	0.830	SVM
		Test	0.860(0.758–0.963)	0.842	0.840	0.844	SVM
	ICG-R15 ≤ 30% vs. ICG-R15>30%	Training	0.997(0.991–1.000)	0.992	1.000	0.991	XGBoost
		Test	0.938(0.824–1.000)	0.965	0.800	0.981	XGBoost

ICG-R15: indocyanine green retention rate at 15 min, AUC: Area under the ROI curve, ACC: Accuracy