Resting-state fMRI identifies three psychosis subtypes spanning diagnostic categories

An observational study published in Molecular Psychiatry examined whether intrinsic brain connectivity could identify biologically defined subgroups across psychotic disorders. Researchers analyzed resting-state fMRI data from 2,103 participants in B-SNIP 1 and 2, deriving a three-component representation of multiscale functional network connectivity. Unsupervised learning identified three Psychosis Imaging Neurosubtypes (PINs) among probands without using clinical or cognitive features to define the groups.

The subtypes crossed DSM diagnostic boundaries. PIN-1 had the greatest cognitive impairment, with reduced cerebellar–subcortical and visual–sensorimotor connectivity and increased visual–subcortical connectivity. PIN-2 had the most preserved cognition, with reduced connectivity between subcortical networks and visual, sensorimotor and higher-cognition networks. PIN-3 showed intermediate cognition and a different combination of increased and decreased connectivity. Among relatives, 55% aligned with their proband’s subtype, reportedly exceeding chance and DSM-based alignment.

Covariate-adjusted analyses found no significant clinical differences between subtypes, although exploratory analyses identified selected differences in symptom measures and age at onset. The imaging groups only partially corresponded to previously described multimodal B-SNIP biotypes. These findings support imaging-based characterization of psychosis heterogeneity, but do not establish a diagnostic or treatment-selection tool. Familial alignment suggests possible heritability rather than proving a genetic basis.