3D-CNN based discrimination of schizophrenia using resting-state fMRI

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Psychiatric Disorders

Background:

• Schizophrenia diagnosis relies on clinical symptoms, lacking objective biomarkers.
• Neuroimaging techniques like fMRI offer potential for objective diagnostic tools.
• Previous studies show altered brain functional connectivity in schizophrenia.

Purpose of the Study:

• To develop and validate a deep learning framework for discriminating schizophrenia patients from healthy controls using resting-state fMRI data.
• To identify specific functional connectivity patterns associated with schizophrenia.
• To assess the potential of the framework as an adjunct screening tool.

Main Methods:

• Utilized resting-state functional MRI (fMRI) data from 144 subjects (72 schizophrenia patients, 72 controls) from a public dataset.
• Employed a three-dimensional Convolutional Neural Network (3D-CNN) deep learning model for classification.
• Extracted features using Independent Component Analysis (ICA) to analyze functional connectivity networks.

Main Results:

• Achieved 98.09% ± 1.01% ten-fold cross-validated classification accuracy (p < 0.001) and an Area Under the Curve (AUC) of 0.9982 ± 0.015.
• Identified significant differences in functional connectivity across resting-state networks between groups.
• Observed prominent visual-frontal network disconnection and increased default mode network connectivity in schizophrenia patients.

Conclusions:

• The developed 3D-CNN framework demonstrates high accuracy in discriminating schizophrenia.
• ICA-derived functional network maps are effective discriminative imaging features.
• Further validation could lead to its use as an assistive tool for initial schizophrenia screening.