Task-related component analysis for functional neuroimaging and application to near-infrared spectroscopy data

Area of Science:

• Neuroimaging
• Signal Processing
• Biophysics

Background:

• Reproducibility is crucial for scientific validity.
• Existing methods like GLM and ICA have limitations in objectivity and interpretation.
• Need for robust methods to analyze complex neuroimaging data.

Purpose of the Study:

• To propose a novel signal processing method, task-related component analysis (TRCA), for extracting reproducible task-related components from neuroimaging data.
• To offer an objective and automated approach for identifying significant components, overcoming limitations of existing methods.
• To demonstrate the applicability and effectiveness of TRCA across different datasets and modalities.

Main Methods:

• TRCA extracts task-related components by maximizing reproducibility during task periods.
• Weights for linear combinations of time courses are optimized using correlation (CorrMax) or covariance (CovMax) maximization.
• Covariance maximization is solved via a Rayleigh-Ritz eigenvalue problem; statistical significance is tested using eigenvalues.

Main Results:

• TRCA successfully identified task-related components in synthetic and NIRS finger-tapping data.
• Two statistically significant components were found: a hemodynamic response and a piece-wise linear time course.
• TRCA demonstrated robustness to data autocorrelation and adaptability for multi-modal integration and artifact removal.

Conclusions:

• TRCA provides an objective, data-driven method for analyzing neuroimaging data, enhancing reproducibility.
• The method offers a systematic approach to component identification and classification, independent of prior assumptions.
• TRCA shows broad applicability in multi-channel biophysical and behavioral measurements, including NIRS data.