Multi-voxel pattern analysis of fMRI data predicts clinical symptom severity

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Multi-voxel pattern analysis (MVPA) is a powerful fMRI technique for healthy populations.
• Its application in patient groups, particularly for symptom prediction, remains underexplored.
• Existing fMRI datasets may hold untapped potential for MVPA research.

Purpose of the Study:

• To investigate MVPA's sensitivity in predicting patient symptoms.
• To demonstrate MVPA's utility with pre-existing fMRI datasets not initially designed for MVPA.
• To explore MVPA as a potential functional biomarker for Autism Spectrum Disorder (ASD) severity.

Main Methods:

• Analyzed fMRI data from individuals with ASD and controls focusing on face processing.
• Applied MVPA to classify patterns of brain activity.
• Utilized univariate analyses and searchlight analyses across the ventral temporal lobes.
• Correlated MVPA classification performance with standardized ASD symptom severity measures.

Main Results:

• MVPA classification performance showed reliable correlations with ASD symptom severity, surpassing univariate measures.
• These correlations remained robust despite variations in region of interest (ROI) definition.
• Searchlight analysis identified specific ventral temporal lobe regions linking MVPA performance to symptom severity, which were missed by mean activation analysis.

Conclusions:

• MVPA demonstrates potential as a sensitive functional biomarker for ASD symptom severity.
• This approach can yield significant insights from existing fMRI datasets.
• MVPA offers a valuable tool for understanding neural underpinnings of patient conditions and severity.