Table 3.

Comparison of performances of five different methods on Experiment 2. The reported values are the correlation coefficient (for regressions of MMSE and ADAS-Cog change) and accuracy, sensitivity and specificity (for MCI-C vs. MCI-NC classification), averaged on 10-fold tests (with standard deviation also reported).

Methods	Correlation coefficient		MCI-C vs. MCI-NC
	MMSE change	ADAS-Cog change	Accuracy	Sensitivity	Specificity
MRI-based	0.419 ± 0.019	0.455 ± 0.037	0.620 ± 0.058	0.566 ± 0.069	0.602 ± 0.056
PET-based	0.434 ± 0.027	0.401 ± 0.046	0.639 ± 0.046	0.570 ± 0.067	0.623 ± 0.069
CSF-based	0.327 ± 0.018	0.425 ± 0.028	0.518 ± 0.086	0.454 ± 0.094	0.493 ± 0.089
CONCAT	0.484 ± 0.009	0.475 ± 0.045	0.654 ± 0.050	0.573 ± 0.062	0.651 ± 0.064
Proposed M3T	0.511 ± 0.021	0.531 ± 0.032	0.739 ± 0.038	0.686 ± 0.051	0.736 ± 0.045

MCI-C = MCI converter, MCI-NC = MCI non-converter, MMSE = Mini-Mental State Examination, ADAS-Cog = Alzheimer’s Disease Assessment Scale - Cognitive Subscale