Table 4.

Diagnostic Performance of Various Machine Learning Models in Distinguishing IMA from INMA in SPNs ≤3 cm.

Model	Cohorts	AUC	Accuracy	Precision	Recall	F1 score
SVM	Training	0.829	0.835	0.303	0.719	0.426
	Test	0.846	0.660	0.239	0.944	0.382
CART	Training	0.743	0.732	0.200	0.719	0.313
	Test	0.636	0.735	0.245	0.667	0.358
KNN	Training	0.829	0.620	0.172	0.906	0.288
	Test	0.662	0.525	0.168	0.833	0.280
LR	Training	0.826	0.663	0.190	0.906	0.314
	Test	0.809	0.741	0.286	0.889	0.432

Abbreviation: IMA invasive mucinous adenocarcinoma, LADC lung adenocarcinoma, AUC area under the curve, INMA invasive non-mucinous adenocarcinoma, SVM support vector machine, CART classification and regression trees, KNN k-nearest neighbors, LR logistic regression, SPN solitary pulmonary nodule.