Background: Functional MRI offers insight into the functioning of brain networks of patients with psychiatric disorders. Machine learning analysis can be used to create diagnostic models and to predict treatment outcome.
Aim: To provide an overview of recent insights on diagnostic and predictive neuroimaging biomarkers.
Method: Narrative review based on recent literature.
Results: Large-scale studies suggest that diagnostic models for most disorders have limited accuracy. In contrast, meta-analyses of small-scale studies suggest that treatment outcome for depression and psychotic disorders can be predicted well.
Conclusion: This creates the opportunity to develop prediction models that can help practitioners in making a treatment plan and thereby improve treatment outcomes.