From Snapshots to Stable Outcomes: Resting-State Functional Magnetic Resonance Imaging-Based Prognosis of Functioning...

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Early recovery is crucial for favorable outcomes in psychotic and affective disorders.
• Transdiagnostic brain activity patterns may indicate future poor outcomes, enabling timely interventions.

Purpose of the Study:

• To evaluate the transdiagnostic prognostic value of resting-state fMRI fractional amplitude of low-frequency fluctuations (fALFF) for functional outcomes.
• To assess the generalizability of machine learning models across different sites and diagnostic groups.

Main Methods:

• Machine learning models were trained using resting-state fMRI fALFF (slow-5 and slow-4 sub-bands) from patients at clinical high-risk for psychosis (n=217) and recent-onset depression (n=198).
• Leave-site-out cross-validation was employed to assess geographic generalizability.
• Outcomes were defined as functional impairment at 9- or 18-month follow-up.

Main Results:

• Transdiagnostic models predicting stable functioning showed higher balanced accuracy (BAC) than 'snapshot' models.
• Decreased slow-5 fALFF in specific brain networks predicted impairment with 67% BAC, generalizing to recent-onset psychosis.
• Slow-5 fALFF models significantly improved prognostic accuracy over expert ratings for both disability and symptom domains.

Conclusions:

• fALFF demonstrates significant prognostic value for functional impairment in individuals at risk for psychosis and those with early depression.
• Candidate imaging biomarkers were identified to enhance prognostication, supporting personalized prevention and recovery strategies.