Table 2.

Three-fold cross-validation performance of multimodal prediction models using radiomics, data fusion strategy, and our proposed MMIDFNet methods.

Models	Fold	ACC	AUC	SEN	SPE	PPV	NPV	Kappa
Radiomics	1	0.851	0.874	0.702	0.885	0.905	0.945	0.699
	2	0.824	0.875	0.705	0.897	0.793	0.922	0.672
	3	0.836	0.862	0.706	0.914	0.724	0.927	0.695
	Average	0.837	0.870	0.704	0.899	0.807	0.931	0.689
	95% CI	[0.815,0.859]	[0.859,0.882]	[0.701, 0.708]	[0.875,0.922]	[0.661,0.954]	[0.912,0.951]	[0.665,0.712]
Data fusion	1	0.865	0.898	0.732	0.913	0.890	0.943	0.740
	2	0.838	0.879	0.744	0.926	0.741	0.922	0.713
	3	0.836	0.871	0.717	0.908	0.745	0.921	0.695
	Average	0.846	0.883	0.731	0.916	0.792	0.929	0.716
	95% CI	[0.820,0.872]	[0.860,0.905]	[0.709, 0.753]	[0.901,0.931]	[0.656,0.928]	[0.909,0.949]	[0.680,0.752]
Decision fusion	1	0.892	0.902	0.781	0.919	0.924	0.959	0.789
	2	0.865	0.909	0.741	0.926	0.821	0.949	0.749
	3	0.877	0.896	0.795	0.946	0.842	0.939	0.780
	Average	0.878	0.902	0.772	0.930	0.862	0.949	0.773
	95% CI	[0.856,0.900]	[0.892,0.913]	[0.727, 0.817]	[0.908,0.953]	[0.775,0.949]	[0.933,0.965]	[0.739,0.806]

ACC, accuracy; SEN, sensitivity; SPE, specificity; PPV, positive predictive value; NPV, negative predictive value; CI, confidence interval; Bold Value, average value of 3 folds.