Multi-classification model incorporating radiomics and clinic-radiological features for predicting invasiveness and differentiation of pulmonary adenocarcinoma nodules

Table 3 The performance of ML models in testing cohort of two-classification model

Cohort	ML model	Clinic-radiological model				Radiomics model				Combined model
		AUC	Accuracy	Sensitivity	Specificity	AUC	Accuracy	Sensitivity	Specificity	AUC	Accuracy	Sensitivity	Specificity
Train	LR	0.907	0.828	0.821	0.836	0.975	0.914	0.891	0.939	0.972	0.891	0.875	0.909
	SVM	0.932	0.874	0.864	0.885	0.962	0.877	0.848	0.909	0.985	0.934	0.924	0.945
	KNN	0.918	0.848	0.815	0.885	0.967	0.900	0.902	0.897	0.972	0.903	0.864	0.945
	DT	0.900	0.837	0.793	0.885	0.943	0.903	0.880	0.927	0.966	0.937	0.967	0.903
	RF	0.891	0.819	0.799	0.842	0.968	0.900	0.886	0.915	0.961	0.891	0.891	0.891
	GBDT	0.913	0.847	0.900	0.788	0.955	0.874	0.821	0.933	0.960	0.905	0.880	0.933
Test	LR	0.905	0.815	0.813	0.817	0.937	0.854	0.800	0.915	0.934	0.868	0.863	0.873
	SVM	0.905	0.848	0.838	0.859	0.938	0.868	0.825	0.915	0.942	0.894	0.875	0.915
	KNN	0.843	0.781	0.725	0.845	0.925	0.841	0.838	0.845	0.933	0.881	0.863	0.901
	DT	0.868	0.788	0.688	0.901	0.848	0.834	0.825	0.845	0.815	0.781	0.775	0.789
	RF	0.909	0.828	0.800	0.859	0.927	0.848	0.863	0.831	0.940	0.881	0.900	0.859
	GBDT	0.905	0.836	0.858	0.812	0.932	0.861	0.788	0.944	0.941	0.867	0.837	0.901

Bold values indicated the best performance in the two-classification task
ML machine learning, LR logistic regression, SVM support vector machine, KNN K-nearest neighbor, DT decision tree, RF random forest, GBDT gradient boosting decision tree, AUC area under the curve

ISSN: 1475-925X