Table 7.

MCI/NC classification comparison of our proposed method (iMSF) and missing value estimation methods (Zero, KNN, SVD and EM) in terms of accuracy, sensitivity and specificity when the training percentage varies from 1/2 to 3/4 as well as leave-one-out (LOO). In this experiment, we used the full multi-source data including MRI, PET, proteomics and CSF with 608 subjects in total.

Accuracy	Training Size	iMSF	Zero	EM	KNN	SVD
	50.0%	0.8872	0.8346	0.8095	0.8142	0.8115
	66.7%	0.9033	0.843	0.8132	0.8113	0.8193
	75.0%	0.8927	0.8462	0.8088	0.8106	0.8165
	LOO	0.9116	0.8481	0.8204	0.8287	0.8204
Sensitivity	Training Size	iMSF	Zero	EM	KNN	SVD
	50.0%	0.6228	0.2537	0.1573	0.174	0.1678
	66.7%	0.6922	0.3055	0.1726	0.1768	0.2092
	75.0%	0.7162	0.3184	0.1709	0.1741	0.2093
	LOO	0.7375	0.3125	0.2125	0.2625	0.2
Specificity	Training Size	iMSF	Zero	EM	KNN	SVD
	50.0%	0.9907	0.9955	0.9901	0.9915	0.9898
	66.7%	0.9934	0.9956	0.9931	0.9895	0.9906
	75.0%	0.9949	0.9982	0.9907	0.9912	0.9893
	LOO	0.9965	1	0.9929	0.9894	0.9965