Classification of schizophrenia using general linear model and support vector machine via fNIRS

Area of Science:

• Neuroscience
• Medical Imaging
• Psychiatry

Background:

• Schizophrenia diagnosis relies heavily on subjective clinical experience, increasing the risk of misdiagnosis.
• Objective physiological data and standardized analysis methods are lacking for identifying schizophrenia-related brain patterns.
• Functional near-infrared spectroscopy (fNIRS) offers a potential non-invasive tool for brain activity measurement.

Purpose of the Study:

• To develop and validate an optimized data processing and analysis pipeline for identifying schizophrenia biomarkers using fNIRS.
• To establish a robust set of fNIRS pattern features capable of discriminating between individuals with schizophrenia and healthy controls.
• To improve the objectivity and accuracy of schizophrenia detection.

Main Methods:

• An optimized data-preprocessing method was employed.
• General linear model (GLM) techniques were used for feature extraction.
• Independent sample t-tests were applied for feature selection.
• Support vector machine (SVM) classification was utilized for discrimination.
• Leave-one-out cross-validation was performed for accuracy assessment.

Main Results:

• The combined approach effectively identified schizophrenia-related fNIRS patterns.
• A classification accuracy of 89.5% was achieved in distinguishing schizophrenia patients from healthy individuals.
• The developed method demonstrated robustness in discriminating between the two groups.

Conclusions:

• The proposed integrated method of data preprocessing, feature extraction, feature selection, and SVM classification is effective for identifying schizophrenia.
• fNIRS patterns, when analyzed with this optimized pipeline, can serve as a reliable biomarker for schizophrenia.
• This approach has the potential to reduce misdiagnosis rates and aid in clinical decision-making for schizophrenia.