A machine learning system detects greater neuroanatomic variation among adolescents with a clinically-high risk for future psychosis at an early age, according to results published in JAMA Psychiatry.
These results are consistent with the idea that age-related differing intercepts and trajectories in brain patterns contribute to the heterogeneity of the onset and course of schizophrenia.
The study included participants with a clinically high risk for future psychosis (n=275) and healthy controls (n=109) age 12 to 21 years from the North American Prodrome Longitudinal Study 2. The researchers developed a neuroanatomic-based age prediction model using a supervised machine learning technique with T1-weighted magnetic resonance imaging scans of 953 healthy controls age 3 to 21 years from the Pediatric Imaging, Neurocognition, and Genetics study. This model was then applied to participants of the current study. The researchers used the scans to identify discrepancies between neuroanatomic-based predicted brain age and chronologic age.
When the researchers applied the Pediatric Imaging, Neurocognition, and Genetics-derived model to the 109 healthy controls, it estimated participants’ chronologic ages accurately, providing external validation for the model.
Compared with controls, the clinically high-risk participants had a significantly greater mean gap between model-predicted age and chronologic age (0.64 years; P =.008). The predicted brain age systematically overestimated the chronologic age of clinically high-risk participants in whom psychosis developed between ages 12 and 17 but not for those in whom psychosis developed between the ages of 18 and 21 years.
The researchers found that greater brain age deviation was associated with an increased risk for psychosis and unstable functioning over time in younger clinically high-risk participants (P =.01).
“This pattern of findings is consistent with the notion that contemporaneous neuroanatomic vulnerability is characteristic of individuals expressing prodromal symptoms in early adolescence and imply that a greater brain age gap may be useful in prediction of early-onset forms of psychosis,” the researchers wrote.
Chung Y, Addington J, Bearden CE, et al. Use of machine learning to determine deviance in neuroanatomical maturity associated with future psychosis in youths at clinically high risk [published online July 3, 2018]. JAMA Psychiatry. doi:10.1001/jamapsychiatry.2018.1543