Table 2.

Predictive performance of different methods among MCI groups.

Classifier	Predictive method	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC
SVM	ROI uptake	74.8 ± 2.41	82.2 ± 1.72	66.7 ± 2.51	0.829 ± 0.035
	MCI pattern	76.7 ± 2.48	83.7 ± 4.46	67.1 ± 6.29	0.831 ± 0.026
	Connectome	85.2 ± 2.34	88.1 ± 3.17	81.2 ± 4.28	0.933 ± 0.014
LR model	ROI uptake	72.4 ± 2.73	81.1 ± 5.99	60.7 ± 4.83	0.748 ± 0.037
	MCI pattern	74.8 ± 4.36	82.3 ± 2.49	66.8 ± 5.91	0.829 ± 0.036
	Connectome	82.3 ± 3.29	80.9 ± 3.14	84.3 ± 6.64	0.867 ± 0.043
Random forest	ROI uptake	70.8 ± 4.73	81.1 ± 3.75	59.3 ± 6.23	0.725 ± 0.045
	MCI pattern	73.1 ± 4.02	85.4 ± 2.86	61.4 ± 8.84	0.787 ± 0.032
	Connectome	76.2 ± 3.19	87.6 ± 2.99	62.9 ± 7.48	0.807 ± 0.031

The predictive performance of MCI participants was not involved in the training dataset.