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

Table 5 The performance of ML models in testing cohort of five-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	0.500	0.484	0.491	0.485	0.597	0.592	0.595	0.591	0.631	0.623	0.626	0.623
	SVM-OVR	0.271	0.184	0.267	0.170	0.666	0.667	0.670	0.676	0.426	0.426	0.446	0.438
	LR-OVO	0.509	0.486	0.488	0.501	0.626	0.618	0.618	0.624	0.683	0.670	0.670	0.676
	SVM-OVO	0.557	0.539	0.534	0.613	0.700	0.696	0.685	0.722	0.703	0.693	0.690	0.715
	DT	0.523	0.488	0.491	0.605	0.651	0.624	0.635	0.714	0.663	0.644	0.649	0.696
	KNN	0.574	0.566	0.554	0.625	0.637	0.627	0.622	0.645	0.609	0.612	0.600	0.635
	RF	0.537	0.503	0.512	0.531	0.620	0.613	0.611	0.620	0.689	0.683	0.674	0.705
	GBDT	0.523	0.462	0.491	0.545	0.694	0.685	0.677	0.732	0.700	0.692	0.686	0.722
Test	LR-OVR	0.420	0.416	0.427	0.415	0.467	0.440	0.475	0.429	0.520	0.507	0.531	0.500
	SVM-OVR	0.287	0.191	0.283	0.166	0.500	0.478	0.502	0.474	0.427	0.416	0.445	0.421
	LR-OVO	0.453	0.446	0.452	0.448	0.493	0.462	0.483	0.453	0.547	0.528	0.550	0.522
	SVM-OVO	0.513	0.506	0.505	0.514	0.553	0.531	0.536	0.559	0.607	0.594	0.603	0.593
	DT	0.500	0.455	0.467	0.539	0.480	0.448	0.471	0.460	0.500	0.490	0.494	0.502
	KNN	0.427	0.417	0.407	0.462	0.453	0.438	0.444	0.446	0.460	0.441	0.443	0.442
	RF	0.467	0.429	0.454	0.423	0.507	0.497	0.500	0.505	0.527	0.499	0.507	0.513
	GBDT	0.440	0.378	0.419	0.412	0.493	0.457	0.479	0.480	0.500	0.460	0.484	0.478

Bold values indicated the best performance in the five-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