Towards Automated Classification of Visual Hallucination Presence in Psychosis using Resting-State fMRI.

Area of Science:

• Neuroimaging
• Psychiatric Disorders
• Machine Learning

Background:

• Visual hallucinations (VH) significantly impair quality of life and correlate with psychosis severity.
• Identifying reliable imaging biomarkers for VH is crucial for diagnosis and treatment.

Purpose of the Study:

• To develop and validate an automated pipeline for classifying visual hallucinations using resting-state fMRI.
• To compare the effectiveness of different functional connectivity features for VH classification.
• To identify potential imaging biomarkers for visual hallucinations.

Main Methods:

• Utilized resting-state functional MRI data from individuals with and without visual hallucinations.
• Assessed five functional connectivity features: Regional Homogeneity, Voxel-Mirrored Homotopic Connectivity, Amplitude of Low Frequency Fluctuations, Fractional Amplitude of Low Frequency Fluctuations, and Eigenvector Centrality Mapping.
• Employed Pearson correlation for feature selection and a linear support vector machine for classification, validated on 45 participants.

Main Results:

• The automated pipeline demonstrated performance in classifying visual hallucinations based on fMRI data.
• Evaluated classification accuracy, sensitivity, specificity, and feature weight interpretability.
• The study identified key functional connectivity features relevant for VH classification.

Conclusions:

• The developed pipeline offers a valuable tool for identifying imaging biomarkers of visual hallucinations.
• Biomarker-based classification can potentially enhance clinical measures, predict psychosis progression, and guide personalized schizophrenia treatment.
• Findings provide insights into VH classification and underlying neural mechanisms.