A new blind source separation framework for signal analysis and artifact rejection in functional Near-Infrared...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Artifact rejection in functional Near-Infrared Spectroscopy (fNIRS) is critical for real-world data analysis, yet lacks a standardized approach.
• Existing methods often struggle with complex signal characteristics like non-instantaneous coupling and correlated noise.
• Blind-Source-Separation (BSS) methods are rarely applied despite their potential due to these challenges.

Purpose of the Study:

• To develop and validate a novel BSS framework for robust artifact rejection in fNIRS signals.
• To enable the analysis of fNIRS data from freely moving subjects under cognitive workload.
• To address limitations of current artifact removal techniques in complex environments.

Main Methods:

• Integration of Independent Component Analysis (ICA) exploiting higher-order statistics and temporal dependencies.
• Multimodal approach combining fNIRS with accelerometer signals.
• Application of Canonical Correlation Analysis (CCA) with temporal embedding.
• Implementation of the Blind Source Separation and Accelerometer based Artifact Rejection and Detection (BLISSA^2RD) method.

Main Results:

• BLISSA^2RD significantly reduces movement-induced artifacts by up to two orders of magnitude across 17 subjects.
• The method improves the Signal-to-Noise Ratio (SNR) of hemodynamic signals by up to 10dB.
• Outperforms conventional PCA, Spline, and Wavelet-based methods in extracting simulated Hemodynamic Response Functions from contaminated data.

Conclusions:

• The proposed multimodal BSS framework offers a powerful solution for artifact rejection in fNIRS.
• BLISSA^2RD enables reliable analysis of cognitive workload in naturalistic settings with freely moving individuals.
• This work provides a blueprint for advanced multivariate signal analysis in fNIRS and beyond.