Diagnosis of Schizophrenia Based on Deep Learning Using fMRI

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Schizophrenia is a prevalent brain disorder in young individuals.
• Early diagnosis and treatment are crucial for reducing societal and familial burdens.
• Objective diagnostic markers for schizophrenia are currently lacking.

Purpose of the Study:

• To enhance the classification accuracy of magnetic resonance data for schizophrenia.
• To develop an objective method for diagnosing schizophrenia using functional magnetic resonance imaging (fMRI).
• To address challenges in classifying small samples and high-dimensional neuroimaging data.

Main Methods:

• Utilized functional magnetic resonance imaging (fMRI) data from schizophrenia patients and healthy controls.
• Applied convolutional neural network (CNN) algorithms, specifically the VGG16 architecture.
• Employed transfer learning for feature extraction and correlation analysis on regions of interest.
• Classified functional connectivity patterns between schizophrenia and control groups.

Main Results:

• Achieved a classification accuracy of up to 84.3% for schizophrenia using VGG16-based fMRI analysis.
• Demonstrated the effectiveness of the proposed method in classifying high-dimensional fMRI data.
• Showcased improved generalization capabilities of deep learning models for neuroimaging analysis.

Conclusions:

• The VGG16-based fMRI classification method shows significant potential for improving early schizophrenia diagnosis.
• This approach effectively addresses the challenges of small sample sizes and high dimensionality in neuroimaging datasets.
• The study highlights the utility of deep learning in advancing objective diagnostic tools for psychiatric disorders.