Table 3.
Average specificity of cross-validation results, feature vector size per columns. Rows correspond to scalar feature mappings.
| Measure | Feat.Map. | HG | 500 | 1000 | 5000 | 10000 | |
|---|---|---|---|---|---|---|---|
| SZAH vs. SZNAH | FC | OS-LAAM | L | 95 | 95 | 95 | 85 |
| R | 80 | 85 | 90 | 85 | |||
| BF-LAAM | L | 100 | 95 | 90 | 75 | ||
| R | 100 | 100 | 100 | 100 | |||
| LA | ReHo | - | 100 | 100 | 100 | 100 | |
| ALFF | - | 75 | 75 | 85 | 85 | ||
| fALFF | - | 95 | 100 | 100 | 95 | ||
| SZAH vs. HC | FC | OS-LAAM | L | 63.3 | 71.7 | 51.7 | 51.7 |
| R | 70 | 75 | 53.3 | 50 | |||
| BF-LAAM | L | 100 | 100 | 100 | 100 | ||
| R | 100 | 100 | 97.7 | 96.7 | |||
| LA | ReHo | - | 96.7 | 96.6 | 96.7 | 96.7 | |
| ALFF | - | 65 | 58.3 | 46.7 | 51.7 | ||
| fALFF | - | 100 | 100 | 96.7 | 96.7 | ||
| SZNAH vs. HC | FC | OS-LAAM | L | 81.7 | 83.3 | 81.7 | 78.3 |
| R | 90 | 90 | 81.7 | 76.7 | |||
| BF-LAAM | L | 100 | 100 | 100 | 100 | ||
| R | 100 | 100 | 100 | 100 | |||
| LA | ReHo | - | 96.7 | 96.7 | 96.7 | 96.7 | |
| ALFF | - | 90 | 93.3 | 83.3 | 73.3 | ||
| fALFF | - | 100 | 100 | 100 | 100 | ||
| SZ vs. HC | FC | OS-LAAM | L | 55 | 40 | 38.3 | 40 |
| R | 46.7 | 45 | 36.7 | 35 | |||
| BF-LAAM | L | 93.3 | 95 | 83.3 | 76.7 | ||
| R | 96.7 | 90 | 85 | 83.3 | |||
| LA | ReHo | - | 96.7 | 96.7 | 96.7 | 96.7 | |
| ALFF | - | 56.7 | 48.3 | 40 | 41.7 | ||
| fALFF | - | 100 | 100 | 96.7 | 96.7 |
Column HG indicates the left (L) or right (R) Heschl’s Gyrus ROI. Results above 90% are highlighted in bold. Key to abbreviations: FC = Functional connectivity. LA = Local Activity. OS-LAAM = one-sided lattice auto-associative memories. BF-LAAM = background/foreground lattice auto-associative memories. ReHo = regional homogeneity. ALFF = amplitude of low frequency fluctuations. fALFF = fractional amplitude of low frequency fluctuations.