Table 4.

Performance comparison among machine learning models

Data	ICG-R15		SVM	RF	ExtraTrees	XGBoost	LightGBM
MRI	ICG-R15 ≤ 10% vs. ICG-R15>10%	Test set ACC	0.824	0.765	0.824	0.853	0.794
		Test set AUC (95%CI)	0.802(0.639–0.965)	0.839(0.703–0.974)	0.873(0.743–1.000)	0.899(0.784–1.000)	0.806(0.650–0.962)
	ICG-R15 ≤ 20% vs. ICG-R15>20%	Test set ACC	0.824	0.882	0.735	0.824	0.824
		Test set AUC (95%CI)	0.893(0.780–1.000)	0.979(0.941–1.000)	0.878(0.739–1.000)	0.946(0.866–1.000)	0.833(0.632–1.000)
	ICG-R15 ≤ 30% vs. ICG-R15>30%	Test set ACC	0.882	0.618	0.882	0.941	0.794
		Test set AUC (95%CI)	0.922(0.802–1.000)	0.789(0.481–1.000)	0.945(0.866–1.000)	0.961(0.890–1.000)	0.891(0.743–1.000)
CT	ICG-R15 ≤ 10% vs. ICG-R15>10%	Test set ACC	0.772	0.632	0.667	0.842	0.702
		Test set AUC (95%CI)	0.734(0.590–0.879)	0.661(0.514–0.807)	0.723(0.576–0.870)	0.822(0.700–0.944)	0.741(0.610–0.872)
	ICG-R15 ≤ 20% vs. ICG-R15>20%	Test set ACC	0.842	0.667	0.702	0.684	0.684
		Test set AUC (95%CI)	0.860(0.758–0.963)	0.722(0.591–0.853)	0.634(0.478–0.789)	0.709(0.570–0.847)	0.692(0.552–0.832)
	ICG-R15 ≤ 30% vs. ICG-R15>30%	Test set ACC	0.982	0.912	0.807	0.965	0.982
		Test set AUC (95%CI)	0.865(0.600–1.000)	0.871(0.683–1.000)	0.783(0.471–1.000)	0.938(0.824–1.000)	0.925(0.776–1.000)

The performance of the best model is in boldface