Network-based diagnostic probability estimation from resting-state functional magnetic resonance imaging.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomarker Discovery

Background:

• Brain functional connectivity, measured via resting-state fMRI (rs-fMRI), is a key biomarker for diagnosing brain disorders.
• Previous diagnostic models estimated brain networks independently of outcomes, limiting their diagnostic utility.
• A need exists for outcome-aware methods to improve the accuracy of brain disorder diagnosis.

Purpose of the Study:

• To propose a novel regression method integrating supervised sparse hierarchical components analysis (SSHCA) for brain disorder diagnosis.
• To enhance the diagnostic utility of rs-fMRI connectivity data by incorporating outcome-related information.
• To develop a model for accurate prediction of disease states using brain network characteristics.

Main Methods:

• Developed SSHCA, a method with a hierarchical structure for network and scoring models.
• Utilized a regression model, specifically multiple logistic regression, with super scores from SSHCA as predictors.
• Applied the method to both simulated data and real rs-fMRI data for validation.

Main Results:

• The proposed SSHCA method demonstrated high accuracy in predicting diseases in both simulation and real data applications.
• Outcome-related network connections and sub-network scores derived from SSHCA provided valuable interpretability.
• The supervised approach significantly improved the diagnostic power compared to previous outcome-independent methods.

Conclusions:

• The developed SSHCA-based regression method offers a powerful and interpretable approach for diagnosing brain disorders using rs-fMRI data.
• This outcome-aware analysis of brain functional connectivity enhances predictive accuracy.
• SSHCA represents a significant advancement in leveraging neuroimaging biomarkers for clinical diagnosis.