JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Advanced design and classification of wearable near-infrared spectroscopy device using temporal channel reconfiguration multi-graph convolutional neural networks for motor activity

  • 0Department of Electronics and Communication Engineering, SRM Institute of Science and Technology, Vadapalani, India.
Computer Methods in Biomechanics and Biomedical Engineering +

|

June 12, 2025

|

June 12, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

This study introduces a new wearable near-infrared spectroscopy device (WNISD) using temporal channel reconfiguration multi-graph convolutional neural networks (TRMCNN) for accurate motor activity classification. The WNISD-TRMCNN method significantly outperforms existing techniques in classifying hemoglobin oxygenation levels.

Area Of Science

  • Biomedical Engineering
  • Neuroscience
  • Machine Learning

Background

  • Wearable near-infrared spectroscopy (NIRS) is crucial for non-invasive brain monitoring.
  • Accurate classification of hemodynamic responses (HbO and HbR) is essential for understanding motor activity.
  • Existing methods face challenges with motion artifacts and classification accuracy.

Purpose Of The Study

  • To propose an advanced design for a wearable NIRS device (WNISD).
  • To develop a novel classification method using temporal channel reconfiguration multi-graph convolutional neural networks (TRMCNN).
  • To enhance classification accuracy for oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) during motor activity.

Main Methods

  • Real-time fNIRS data collection.
  • Pre-processing using event-triggered consensus Kalman filtering (ETCKF) to mitigate motion artifacts.
  • Classification of HbO and HbR using TRMCNN, optimized with the Young's double slit experiment optimization algorithm (YDSEOA).

Main Results

  • The proposed WNISD-TRMCNN method demonstrated superior performance.
  • High accuracy, precision, and AUC were achieved in classifying HbO and HbR.
  • The method showed improved processing time compared to existing techniques.

Conclusions

  • The WNISD-TRMCNN offers a robust and accurate solution for wearable NIRS-based motor activity analysis.
  • ETCKF effectively removes motion artifacts, enhancing data quality.
  • YDSEOA optimization further boosts the classification performance of the TRMCNN model.

Keywords: