Table 4.
EP classification performance of our model when introducing different cognitive estimation subtasks, using GM images as input.
| Model with Task | acc | F1 | spe | AUC | ||
|---|---|---|---|---|---|---|
| EP | CLC | CLR | ||||
| ✓ | 71.0 ± 4.3 | 70.1 ± 4.4 | 73.8 ± 5.9 | 66.3 ± 3.9 | ||
| ✓ | ✓ | 71.4 ± 3.7 | 70.2 ± 5.4 | 74.4 ± 5.4 | 67.4 ± 4.5 | |
| ✓ | ✓ | 74.9 ±4.3 | 74.5 ±4.2 | 82.3 ±6.3 | 71.1 ±4.1 | |
| ✓ | ✓ | ✓ | 71.1 ± 3.9 | 71.2 ± 3.8 | 77.0 ± 5.9 | 68.0 ± 4.9 |
Here EP denotes early psychosis classification. The n was set to five for CLC. The bold values indicate the best results of each metric under different task settings.