Neuro-psychiatry subgroup identification and classification (BP/SZ) on BSNIP cohort (test)
61.5Subgroup Balanced Accuracy (B-ACC)CE (upper bound)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CE (upper bound)2026.05 | 61.5 | — | — | |
| GM ROI features + UCSLFeatures=age-corrected Gray Matter Region Based-Morphometry2026.05 | 59 | 65.3 | 52.5 | |
| Deep UCSLInput=end-to-end approach2026.05 | 58.9 | 67.1 | 54.3 | |
| SimCLRTraining set=[SZ+BP] only2026.05 | 57.1 | — | — | |
| SupCon2026.05 | 55 | 65.6 | 45.8 | |
| PCLTraining set=[SZ+BP] only2026.05 | 54.2 | — | — | |
| VAE + UCSL2026.05 | 53.4 | 58.8 | 45.9 | |
| SwAVTraining set=[SZ+BP] only2026.05 | 52.2 | — | — | |
| Deep Cluster - v2Training set=[SZ+BP] only2026.05 | 51.7 | — | — | |
| SCANTraining set=[SZ+BP] only2026.05 | 50.9 | — | — | |
| BYOLTraining set=[SZ+BP] only2026.05 | 50.8 | — | — | |
| BCE + K-Means2026.05 | 50.7 | 65.3 | 42.8 |