Preoperative CT-based deep learning radiomics model to predict lymph node metastasis and patient prognosis in bladder cancer: a two-center study

Insights into Imaging

Table 4 The predictive performance results of radiomics machine learning signatures with SMOTE

Set	Classifier	AUC (95%CI)	ACC	SEN	SPE	PPV	NPV
Training	SVM	0.908 (0.873–0.943)	0.847	0.925	0.769	0.800	0.911
	LR	0.765 (0.711–0.819)	0.714	0.748	0.680	0.700	0.730
	XGBoost	1 (1.000–1.000)	0.997	1.000	0.993	0.993	1.000
	NaiveBayes	0.815 (0.767–0.863)	0.616	0.993	0.238	0.566	0.972
	AdaBoost	0.887 (0.850–0.924)	0.793	0.871	0.714	0.753	0.847
	LightGBM	0.969 (0.951–0.987)	0.915	0.918	0.912	0.912	0.918
	KNN	0.959 (0.941–0.977)	0.820	0.980	0.660	0.742	0.970
	MLP	0.827 (0.780–0.874)	0.718	0.810	0.626	0.684	0.767
	GradientBoosting	0.962 (0.943–0.981)	0.864	0.891	0.837	0.845	0.885
External test	SVM	0.686 (0.486–0.887)	0.833	0.500	0.909	0.556	0.889
	LR	0.727 (0.529–0.925)	0.759	0	0.932	0	0.804
	XGBoost	0.782 (0.641–0.922)	0.815	0.200	0.955	0.500	0.840
	NaiveBayes	0.759 (0.596–0.923)	0.741	0.600	0.773	0.375	0.895
	AdaBoost	0.676 (0.456–0.896)	0.759	0.400	0.841	0.364	0.860
	LightGBM	0.893 (0.769–1.000)	0.852	0.400	0.955	0.667	0.875
	KNN	0.673 (0.523–0.823)	0.593	0.500	0.614	0.227	0.844
	MLP	0.727 (0.575–0.880)	0.759	0.300	0.864	0.333	0.844
	GradientBoosting	0.807 (0.631–0.982)	0.815	0.500	0.886	0.500	0.886

AUC area under the receiver operating characteristic curve, CI confidence interval, ACC accuracy, SEN sensitivity, SPE specificity, PPV positive predictive value, NPV negative predictive value