Identifying Resting-State Multifrequency Biomarkers via Tree-Guided Group Sparse Learning for Schizophrenia...

Area of Science:

• Neuroimaging
• Psychiatric Disorders
• Machine Learning

Background:

• Fractional amplitude of low-frequency fluctuations (fALFF) is a potential biomarker for schizophrenia diagnosis using resting-state fMRI.
• Traditional fALFF analysis uses limited frequency bands (0.01-0.08 Hz), potentially missing complex brain activity variations.
• Existing methods often overlook the influence of brain structure on fALFF data during feature selection.

Purpose of the Study:

• To develop an improved model for classifying schizophrenia by analyzing fALFF across multiple frequency bands.
• To incorporate brain structure constraints into feature selection for more robust schizophrenia detection.
• To enhance diagnostic accuracy by leveraging complementary information from multifrequency band biomarkers.

Main Methods:

• Acquired fALFF data across four distinct frequency bands: slow-5 (0.01-0.027 Hz), slow-4 (0.027-0.073 Hz), slow-3 (0.073-0.198 Hz), and slow-2 (0.198-0.25 Hz).
• Utilized random forest to identify significant brain regions (patches) associated with schizophrenia.
• Applied tree-guided group sparse learning for feature selection, incorporating spatial constraints from identified brain patches.
• Employed multi-kernel learning to integrate features from different frequency bands for classification.

Main Results:

• The proposed multifrequency band analysis achieved a classification accuracy of 91.1% for schizophrenia.
• The model successfully identified significant brain patches related to schizophrenia.
• Combining features across multiple frequency bands improved classification performance compared to traditional methods.

Conclusions:

• Multifrequency band analysis of fALFF provides a more comprehensive assessment of spontaneous brain fluctuations in schizophrenia.
• The developed tree-guided group sparse learning and multi-kernel learning model offers a promising approach for objective schizophrenia diagnosis.
• This approach highlights the importance of considering broader frequency spectrums and structural constraints in neuroimaging-based psychiatric disorder research.