The use of MRI or neurophysiologic tests in psychiatric clinical settings remains premature and limited. For those not familiar with this challenge, we can start with a simple question:
No, not yet. And perhaps this is not the best question.
Psychiatric MRI biomarkers are scientifically valuable, but they are not yet robust enough for routine clinical decision-making. Many neuroimaging findings are promising at the group level, but they still lack sufficient specificity, reproducibility, and individual-level validity for diagnosis or treatment selection. In psychiatry, MRI should not be presented as a diagnostic shortcut. Complex psychiatric conditions will probably never be diagnosed by a single neuroimaging exam alone. They require clinical evaluation, longitudinal observation, and multiple biological, cognitive, behavioral, and contextual signals.
However, this does not mean that MRI has no value.
This is a different question. It is not about diagnosis. It is about patient stratification.
A recent article by Grotzinger et al. (2026), Mapping the genetic landscape across 14 psychiatric disorders, provides an important genetic rationale for this discussion. The authors show that schizophrenia and bipolar disorder form a shared genetic factor, referred to as the SB factor. This factor reflects a major component of common-variant genetic liability shared between the two disorders. The article also reports that this shared SB signal is enriched for genes expressed in excitatory neurons, with particularly relevant findings from fetal brain data and from maturing excitatory neurons.
This does not mean that schizophrenia and bipolar disorder are clinically the same. They are not. They differ in symptoms, course, treatment decisions, prognosis, and lived experience.
The point is more precise: schizophrenia and bipolar disorder may share part of their early biological liability, especially at the level of common genetic risk and neurodevelopmentally relevant neuronal biology. This suggests that early biological trajectories may partially overlap before later clinical divergence becomes clearer.
This should be understood as a hypothesis and a plausible clinical-developmental model, not as a tool for clinical decision-making. I am doing science here, with one cautious toe in the water of clinical trial development.
In this model, genetics provides one layer of information: inherited liability. MRI provides another layer: how this liability may be expressed in brain development, structure, connectivity, maturation, and network organization. Together, they may help us ask whether patients who look similar clinically — especially in early psychosis or severe mental illness — actually belong to different biological trajectory groups.
This is especially relevant because psychiatric diagnoses are often unstable or incomplete in early stages. A first psychotic episode may later evolve toward a schizophrenia-spectrum disorder, bipolar disorder, schizoaffective disorder, recurrent affective psychosis, or another clinical trajectory. At that stage, diagnosis alone may not be the best stratification tool for research or clinical trials.
The goal, therefore, is not to replace diagnosis. The goal is to reduce biological heterogeneity within diagnostic categories and across overlapping clinical presentations.
A genetics–MRI stratification framework could help classify patients by biological trajectory rather than diagnosis alone. For example, one subgroup may show stronger schizophrenia-weighted genetic liability combined with MRI markers of neurodevelopmental deviation or network disruption. Another subgroup may show stronger bipolar-weighted liability with a different brain-expression profile. A third group may show broader transdiagnostic liability. These possibilities remain to be tested.
A combined genetics–MRI framework may stratify early psychosis and schizophrenia–bipolar spectrum patients into biologically meaningful trajectory groups that are more informative for prognosis, treatment-response research, and trial enrichment than diagnosis alone.
This is a patient-stratification strategy, not a diagnostic test. Its value would be mainly in research and drug-development contexts: trial enrichment, subgroup discovery, prognosis, and treatment-response prediction. Neuroimaging would remain supportive evidence, not decisive evidence. Genetics would not determine destiny. But together, these layers may help identify more coherent patient subgroups and improve the biological logic of CNS clinical trials.
In short, the question is not whether MRI can distinguish schizophrenia from bipolar disorder today. The better question is whether genetics and MRI can help us understand when these disorders share biological roots, when their trajectories begin to diverge, and whether those trajectories can inform more precise patient stratification in future psychiatric research.