Optimizing functional near-infrared spectroscopy (fNIRS) channels for schizophrenic identification during a verbal...

Area of Science:

• Neuroscience
• Medical Imaging
• Biomedical Engineering

Background:

• Schizophrenia diagnosis often relies on complex neuroimaging techniques.
• Functional near-infrared spectroscopy (fNIRS) offers a portable and non-invasive method for brain activity assessment.
• Reducing the complexity of fNIRS systems can enhance their accessibility and clinical utility.

Purpose of the Study:

• To simplify a 52-channel functional near-infrared spectroscopy (fNIRS) system for improved schizophrenia discrimination during a verbal fluency task (VFT).
• To evaluate the effectiveness of metaheuristic optimization algorithms in feature extraction for fNIRS data.
• To assess the diagnostic accuracy of a reduced-channel fNIRS system compared to a full system.

Main Methods:

• Collected fNIRS data from 100 schizophrenia patients and 100 healthy controls during a VFT.
• Extracted and optimized time average, functional connectivity, and wavelet features using genetic algorithm (GA), particle swarm optimization (PSO), and hybrid algorithms.
• Utilized Support Vector Machine (SVM) for classification and evaluated performance using ten-fold cross-validation.

Main Results:

• Genetic algorithm (GA) and GA-dominant algorithms outperformed PSO algorithms in accuracy.
• An optimal accuracy of 87.00% was achieved using 16 channels with GA and wavelet analysis.
• A parallel hybrid algorithm achieved 86.50% accuracy with 8 channels using time-domain features, comparable to the 52-channel system.

Conclusions:

• A simplified fNIRS system can maintain diagnostic accuracy for schizophrenia, comparable to more complex systems.
• Optimized feature extraction using evolutionary algorithms, particularly on time-domain features, is effective for reducing channel count.
• This simplification promotes wider application of fNIRS in schizophrenia diagnosis, especially in resource-limited settings.