TABLE 4.
Classification performance of ASD identification achieved by six different methods on four datasets (i.e., OHSU, NYU, USM and UCLA) with rs-fMRI data.
| Site | Method | Accuracy (%) | Sensitivity (%) | Specificity (%) | Precision (%) |
| OHSU | SVM | 53.8 ± 5.2 | 55.1 ± 6.1 | 48.9 ± 7.2 | 52.6 ± 6.7 |
| SVM-ATM | 70.9 ± 3.7 | 69.9 ± 5.6 | 66.8 ± 4.1 | 70.1 ± 5.5 | |
| MLP | 64.0 ± 4.5 | 56.5 ± 3.9 | 61.6 ± 4.2 | 60.3 ± 4.7 | |
| Autoencoder | 74.0 ± 3.5 | 66.6 ± 2.9 | 75.5 ± 4.7 | 71.5 ± 4.1 | |
| BLS | 75.5 ± 5.1 | 66.3 ± 3.8 | 72.6 ± 4.9 | 75.3 ± 3.9 | |
| Ours | 80.8 ± 3.8 | 76.9 ± 7.7 | 84.6 ± 7.7 | 83.3 ± 8.3 | |
| NYU | SVM | 57.1 ± 2.5 | 50.3 ± 3.5 | 62.2 ± 2.7 | 57.8 ± 3.9 |
| SVM-ATM | 71.2 ± 5.1 | 53.3 ± 4.2 | 81.0 ± 1.2 | 69.1 ± 6.5 | |
| MLP | 64.3 ± 4.2 | 68.4 ± 3.7 | 60.6 ± 3.9 | 57.1 ± 4.3 | |
| Autoencoder | 65.7 ± 3.2 | 68.8 ± 2.6 | 63.2 ± 2.7 | 61.1 ± 4.8 | |
| BLS | 69.7 ± 3.5 | 67.4 ± 6.3 | 71.1 ± 1.1 | 70.8 ± 1.4 | |
| Ours | 71.4 ± 5.7 | 75.0 ± 1.5 | 68.4 ± 9.4 | 66.7 ± 9.8 | |
| USM | SVM | 64.7 ± 5.1 | 60.6 ± 1.9 | 66.9 ± 5.1 | 60.7 ± 3.9 |
| SVM-ATM | 69.6 ± 4.6 | 44.3 ± 3.8 | 68.2 ± 6.3 | 61.8 ± 4.3 | |
| MLP | 64.1 ± 4.1 | 61.2 ± 3.8 | 65.4 ± 4.2 | 62.9 ± 3.8 | |
| Autoencoder | 62.5 ± 2.8 | 60.0 ± 3.2 | 66.3 ± 4.5 | 62.5 ± 4.1 | |
| BLS | 76.9 ± 3.1 | 78.5 ± 2.9 | 79.8 ± 3.9 | 82.2 ± 3.9 | |
| Ours | 81.6 ± 2.9 | 84.8 ± 4.1 | 76.0 ± 0.9 | 86.6 ± 0.3 | |
| UCLA | SVM | 65.1 ± 5.7 | 68.3 ± 3.5 | 60.8 ± 4.7 | 65.2 ± 3.3 |
| SVM-ATM | 72.2 ± 3.1 | 73.8 ± 4.1 | 68.9 ± 3.8 | 69.2 ± 2.8 | |
| MLP | 71.9 ± 3.5 | 72.7 ± 2.4 | 64.8 ± 3.1 | 66.1 ± 3.2 | |
| Autoencoder | 57.7 ± 4.6 | 68.2 ± 4.1 | 47.4 ± 4.8 | 58.5 ± 3.9 | |
| BLS | 73.2 ± 2.8 | 76.4 ± 4.5 | 65.8 ± 4.9 | 71.6 ± 3.1 | |
| Ours | 79.5 ± 3.1 | 77.8 ± 3.7 | 75.0 ± 2.3 | 79.2 ± 2.3 |