Classification of schizophrenia patients based on resting-state functional network connectivity

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Automatic classification of mental disorders using neuroimaging data is a growing area of interest.
• Schizophrenia classification presents challenges due to small datasets and high-dimensional data.
• Previous studies primarily used structural MRI, diffusion tensor imaging, and task-based fMRI, with limited use of resting-state data.

Purpose of the Study:

• To investigate the utility of resting-state functional network connectivity (FNC) for discriminating schizophrenia patients from healthy controls.
• To evaluate the performance of various linear and non-linear classification methods for schizophrenia detection.
• To establish a baseline for using FNC in automated mental disorder classification.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (fMRI) data.
• Extracted functional network connectivity (FNC) features from resting-state data.
• Trained and evaluated several linear and non-linear machine learning classifiers, including k-nearest neighbors (KNNs), on separate training and testing datasets.

Main Results:

• High accuracy was achieved in classifying schizophrenia using resting-state FNC data.
• Non-linear discriminative methods, specifically k-nearest neighbors (KNNs), demonstrated promising results.
• Detailed comparisons of classifier performance and statistical analysis of individual features were conducted.

Conclusions:

• Resting-state functional network connectivity (FNC) is a viable and effective biomarker for schizophrenia classification.
• Machine learning approaches, especially KNNs, are suitable for developing accurate automated diagnostic tools for mental disorders.
• This study represents the first use of FNC features for schizophrenia classification, opening new avenues for research.