“`html
Predicting Response to rTMS in Patients with Schizophrenia
The effectiveness of repetitive transcranial magnetic stimulation (rTMS) as a treatment for schizophrenia varies among patients. This variability can be better understood by using artificial intelligence to analyze MRI, clinical, sociodemographic, and genetic data. We developed and tested rTMS response prediction models in patients with schizophrenia from the multisite RESIS trial.
Key Findings
- Our predictive models achieved a high accuracy of 94% in identifying responders and non-responders in the active-treated group, compared to 50% in the sham-treated group.
- Combining clinical and polygenic risk score (PRS) data increased the accuracy of prediction to 76%, while structural MRI-based classifiers yielded an accuracy of 80%.
- Specific factors such as apparent sadness, inability to feel, educational attainment PRS, and unemployment were found to be most predictive of non-response in the clinical + PRS model. Meanwhile, reductions in grey matter density in certain brain networks were predictive in the structural MRI model.
- The findings suggest that rTMS responders may have higher levels of brain grey matter in certain networks, making them more likely to benefit from rTMS treatment.
Our sequential modelling approach improves predictive performance while minimizing the burden in the clinical setting. The future clinical use of these models will require replication on an international scale using stratified clinical trial designs.
“`
