Hilbert Spectrum-Based Approach for fNIRS Interhemispheric Functional Connectivity Analysis.

Abstract

Functional Near-Infrared Spectroscopy (fNIRS) is a promising neuroimaging technique due to its non-invasive nature and tolerance to movement. However, its interpretation is often hindered by physiological and systemic artifacts that mask neural activity. In this work, we propose a data-driven framework to extract neural components from fNIRS signals and to analyze interhemispheric functional connectivity in language disorders. The approach is based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) combined with Hilbert spectral analysis to isolate oscillatory components associated with neural hemodynamics without relying on predefined frequency filters. Then, functional connectivity is computed between channels across hemispheres and statistically significant connections are identified through Mann–Whitney U test and subsequently used as features within a nested classification framework based on XGBoost. The results reveal altered interhemispheric connectivity patterns and demonstrate the discriminative potential of the extracted features, achieving stable classification performance across cross-validation folds.