For a study, researchers sought to use clinical and brain structural imaging data from early psychosis and depression phases and apply machine learning approaches to cluster, compare, and combine subgroup answers.
Between February 1, 2014, and July 1, 2019, a referred patient sample of those with high clinical risk for psychosis (CHR-P), recent-onset psychosis (ROP), recent-onset depression (ROD), and healthy controls were recruited for a multisite, naturalistic, longitudinal cohort study (10 sites in 5 European countries; including major follow-up intervals at 9 and 18 months). Between January 2020 and January 2022, data were evaluated.
A nonnegative matrix factorization approach dissected clinical (287 variables) and parcellated brain structural volume (204 gray, white, and cerebrospinal fluid regions) data from the CHR-P, ROP, ROD, and healthy controls research groups independently. The cluster number was selected using stability criteria using layered cross-validation. Validation goals were compared between subgroup solutions (premorbid, longitudinal, and schizophrenia polygenic risk scores). Multiclass supervised machine learning yielded a transferrable solution to the validation sample.
The discovery group had 749 people, whereas the validation group included 610 people. Individuals with CHR-P (n=287), ROP (n=323), ROD (n=285), and healthy controls (n=464) were included. The average (standard deviation) age was 25.1 (5.9) years, and 702 (51.7%) were female. A clinical four-dimensional solution distinguished people based on positive symptoms, negative symptoms, depression, and functioning, revealing relationships with all validation goals. Brain clustering identified a subgroup with dispersed brain volume decreases related to negative symptoms, worse performance IQ, and higher polygenic risk scores for schizophrenia. The multilevel findings discriminated between normal and disease-related brain changes. The external sample corroborated the subgroup results considerably.
The findings of the longitudinal cohort research gave stratifications that went beyond the manifestation of positive symptoms and span sickness stages and diagnoses. Clinical findings highlighted the significance of negative symptoms, depression, and functioning. The brain findings indicated significant overlap across sickness phases and normative variance, which may indicate a characteristic vulnerability independent of individual presentations. Premorbid, longitudinal, and genetic risk validation revealed that the subgroups were clinically important for preventative therapies.