Related Experiment Video
Updated: Sep 11, 2025
05:59
Author Spotlight: Unlocking New Insights in fNIRS Studies - A Novel Framework for Inter-Brain Synchrony Analysis
Published on: October 6, 2023
2.7K
Schizophrenia Recognition Based on Gramian Angular Field Combining Activation Features: A Functional Near-Infrared
IEEE Journal of Biomedical and Health Informatics
|August 12, 2025
Summary
This study introduces a new method using Gramian Angular Field (GAF) images with functional near-infrared spectroscopy (fNIRS) to improve schizophrenia (SCZ) diagnosis. The GAF approach enhances accuracy in identifying SCZ compared to traditional methods.
Area of Science:
- Neuroscience
- Medical Imaging
- Psychiatry
Background:
- Schizophrenia (SCZ) incidence is rising globally, necessitating improved diagnostic tools.
- Functional near-infrared spectroscopy with verbal fluency task (fNIRS-VFT) offers objective neurofunctional assessment but faces challenges in feature extraction and utilization.
- Existing methods struggle with insufficient time series feature extraction, inadequate feature utilization, and unstable robustness.
Purpose of the Study:
- To develop and evaluate a novel fNIRS classification strategy using Gramian Angular Field (GAF) coding integrated with activation information for enhanced SCZ recognition.
- To compare the proposed GAF-based approach with traditional fNIRS feature extraction methods and other image-based techniques.
Main Methods:
- Designed a fNIRS classification strategy by coding fNIRS signals into 2D Gramian Angular Field (GAF) virtual images, incorporating activation information.
- Processed fNIRS data from 200 participants into virtual images for SCZ recognition.
- Compared classification performance using GAF images with activation information against GAF images without activation, recurrence plots, Markov transition fields, and traditional fNIRS features.
Main Results:
- The GAF image coding approach integrating activation information significantly improved SCZ recognition accuracy by 7.6% over GAF without activation and 4.9% over CNN HbO signal classification.
- ShuffleNetV2 model achieved the highest accuracy of 81.0% on the cross-validation dataset and 72.0% on an external test set.
- The proposed method demonstrated superior performance compared to EfftivenetV2 with GAF images without activation (73.4% accuracy) and CNN HbO signals classification (75.5% accuracy).
Conclusions:
- Gramian Angular Field (GAF) virtual image coding, when integrated with activation information, presents a novel and effective strategy for supporting SCZ screening and diagnosis.
- This approach enhances the application of fNIRS technology in the clinical diagnosis of psychiatric disorders.
- The findings highlight the potential of advanced image coding techniques for objective neurofunctional assessment in psychiatry.
