Brain Model Based on the Canonical Ensemble with Functional MRI: A Thermodynamic Exploration of the Neural System
Table 8
Results of the classification for AD/NC (%).
Feature selection
Distance
Correlation
Cosine
Euclidean
TP-Kendall
Specificity
84.93 ± 2.21
85.18 ± 2.18
79.77 ± 2.41
Sensitivity
73.48 ± 2.45
73.41 ± 2.57
73.20 ± 2.38
Accuracy
80.35 ± 1.71
80.47 ± 1.72
77.14 ± 1.66
TP-Expert
Specificity
73.32 ± 1.30
71.81 ± 1.33
71.73 ± 1.17
Sensitivity
70.91 ± 1.91
67.69 ± 2.03
68.74 ± 1.94
Accuracy
72.36 ± 1.00
70.16 ± 1.02
70.53 ± 1.01
BN-Kendall
Specificity
73.55 ± 2.59
77.99 ± 2.52
75.55 ± 2.60
Sensitivity
65.96 ± 3.81
66.22 ± 3.19
65.62 ± 3.58
Accuracy
70.51 ± 2.04
73.28 ± 1.84
71.58 ± 1.95
TP-Kendall, TP-Expert, and BN-Kendall represent three different methods of feature selection designed in Section 2.5.2. Correlation, cosine, and Euclidean represent correlation distance, cosine distance, and Euclidean distance.