Table 3.

Results for ML models on MOLTEST-BIS test set. The bold value shows the superiority of a particular metric and the set of data between ML approaches. LR—logistic regression, RF—random forest, CI—95% confidence interval.

Metric	Radiomics		Metabolomics		Statistical Integration		Product Integration
	LR	RF	LR	RF	LR	RF	LR	RF
Sensitivity	0.60	0.70	0.55	0.70	0.70	0.75	0.78	0.63
Specificity	0.85	0.70	0.55	0.55	0.80	0.60	0.73	0.69
PPV	0.80	0.70	0.55	0.61	0.78	0.65	0.70	0.75
NPV	0.68	0.70	0.55	0.65	0.73	0.71	0.80	0.55
F1	0.69	0.70	0.55	0.65	0.74	0.70	0.74	0.68
Balanced accuracy	0.73	0.70	0.55	0.63	0.75	0.68	0.75	0.66
AUC (%)	83.0	70.3	55.5	60.3	83.0	73.0	84.8	71.5
AUC 95% CI	70–96	53–87	38–74	42–79	70–96	56–89	73–97	55–88