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

Table 4 The performance of ML models in testing cohort of three-classification model

Cohort	ML model	Clinic-radiological model				Radiomics model				Combined model
		Accuracy	F1-score	Recall	Precision	Accuracy	F1-score	Recall	Precision	Accuracy	F1-score	Recall	Precision
Train	LR-OVR	0.723	0.704	0.737	0.696	0.809	0.781	0.789	0.777	0.837	0.823	0.834	0.816
	SVM-OVR	0.451	0.389	0.461	0.410	0.794	0.773	0.790	0.765	0.723	0.699	0.697	0.702
	LR-OVO	0.737	0.711	0.705	0.721	0.849	0.836	0.825	0.851	0.871	0.869	0.865	0.873
	SVM-OVO	0.749	0.721	0.713	0.734	0.846	0.830	0.820	0.842	0.883	0.878	0.870	0.886
	DT	0.723	0.700	0.704	0.696	0.820	0.743	0.726	0.876	0.820	0.743	0.726	0.876
	KNN	0.734	0.705	0.680	0.777	0.797	0.750	0.733	0.784	0.777	0.741	0.724	0.771
	RF	0.749	0.717	0.695	0.770	0.851	0.822	0.806	0.849	0.851	0.825	0.804	0.866
	GBDT	0.789	0.734	0.714	0.789	0.780	0.721	0.701	0.788	0.763	0.745	0.738	0.753
Test	LR-OVR	0.713	0.672	0.710	0.673	0.740	0.702	0.703	0.707	0.733	0.696	0.697	0.701
	SVM-OVR	0.413	0.333	0.388	0.361	0.713	0.674	0.688	0.670	0.693	0.633	0.630	0.641
	LR-OVO	0.733	0.695	0.687	0.710	0.747	0.708	0.709	0.711	0.747	0.702	0.693	0.715
	SVM-OVO	0.740	0.696	0.697	0.700	0.753	0.712	0.706	0.724	0.767	0.718	0.710	0.731
	DT	0.707	0.680	0.682	0.677	0.653	0.563	0.555	0.669	0.653	0.562	0.555	0.643
	KNN	0.627	0.561	0.545	0.640	0.633	0.572	0.577	0.577	0.673	0.598	0.589	0.634
	RF	0.713	0.660	0.646	0.696	0.707	0.639	0.629	0.672	0.700	0.636	0.629	0.652
	GBDT	0.700	0.635	0.625	0.667	0.713	0.645	0.635	0.678	0.720	0.672	0.652	0.727

Bold values indicated the best performance in the three-classification task
ML machine learning, OVR one versus rest, OVO one versus one, LR logistic regression, SVM support vector machine, KNN K-nearest neighbor, DT decision tree, RF random forest, GBDT gradient boosting decision tree

ISSN: 1475-925X