Figure 9.
The performance (ACC, SPE, and SEN) for MCI identification based on fusion of PCn-based BFN and PC1-based BFN under different p-values. The PCn&1 corresponds to the fused performance of PCn and PC1, n = 2,3,4, conv. The dotted line is the performance of PC1 itself (without fusion with PCn). We can observe that the fusion of low- and higher-order BFNs fails to improve the ACC and SPE in most cases, but generally improves SEN.