Table 3.
Classification accuracy using 5-fold cross-validation on individual data centers using our proposed method, ASD-DiagNet (with and without data augmentation), compared with other methods.
| Site | ASD-DiagNet | ASD-DiagNet (no aug.) | Heinsfeld et al., 2018 | SVM | Random- Forest |
|---|---|---|---|---|---|
| Caltech | 52.8 | 49.9 | 52.3 | 46.9 | 54.2 |
| CMU | 68.5 | 67.4 | 45.3 | 66.6 | 62.4 |
| KKI | 69.5 | 68.6 | 58.2 | 66.4 | 66.6 |
| Leuven | 61.3 | 57 | 51.8 | 59.8 | 59.8 |
| MaxMun | 48.6 | 51.4 | 54.3 | 53.8 | 49.2 |
| NYU | 68 | 65.1 | 64.5 | 71.4 | 61.8 |
| OHSU | 71.9 | 74 | 79.4 | 54.3 | |
| Olin | 65.1 | 58.8 | 44 | 59.5 | 52.2 |
| Pitt | 67.8 | 65.9 | 59.8 | 66.3 | 59.9 |
| SBL | 51.6 | 47.5 | 46.6 | 60 | 48.3 |
| SDSU | 63 | 61.3 | 63.6 | 58.7 | 62.7 |
| Stanford | 64.2 | 53 | 48.5 | 51.4 | 62.1 |
| Trinity | 54.1 | 51.2 | 61 | 53.1 | 54.5 |
| UCLA | 73.2 | 70.3 | 57.7 | 72.1 | 69.3 |
| USM | 68.2 | 65.1 | 62 | 73.2 | 58 |
| UM | 63.8 | 65.7 | 57.6 | 64.2 | 64.8 |
| Yale | 63.6 | 61.7 | 53 | 61.6 | 55.3 |
| Average | 63.8 | 60.7 | 56.1 | 62.6 | 58.6 |
Bold and color values corresponds to highest accuracy achieved among all datasets.