¶
Predictive models¶
We develop MRI-based brain signatures of behaviour and clinical traits — from functional connectivity and task activity to morphology — often with external validation. Alongside the signatures themselves, we build tools that improve statistical power, quantify confounding, and make discovery–validation designs more credible.
Signatures & applications¶
| Resource | One-line summary |
|---|---|
| RPN-signature | Resting-state Pain susceptibility Network: pain-free FC predicts individual pain sensitivity (Nat Commun, 2020). |
| RCPL signature | Externally validated resting-state connectivity signature of pain-related learning (Commun Biol, 2024). |
| CTP / brain morphology | Multicenter ML model: brain morphology predicts individual pain sensitivity (Pain, 2023). |
| PARES | Task-fMRI signature of bidirectional pain change (exacerbation and relief), generalising to chronic back pain. |
Methods & tooling¶
| Resource | One-line summary |
|---|---|
| AdaptiveSplit | Registered models + adaptive sample splitting for credible external validation (GigaScience, 2025). |
| mlconfound | Quantify confounding bias in ML predictions (GigaScience, 2022). |
| pTFCE | Probabilistic threshold-free cluster enhancement for more powerful neuroimaging inference (NeuroImage, 2019). |
Related work also appears under Brain Models, Software, and Publications.
- Li, J., Kincses, B., Schmidt, K., Forkmann, K., Busch, L., Kaur, J., Schlitt-Nguyen, F., Wiech, K., Bingel, U., & Spisak, T. (2026). Beyond pain: a brain-based biomarker predicts individual pain relief. 10.64898/2026.08.21.746216