Decoding Contralateral Finger Movements from fNIRS Signals Using Phase Connectivity Features

Functional near-infrared spectroscopy (fNIRS), an emerging and non-invasive neuroimaging tool, holds great promise for brain-computer interface (BCI) applications and diagnostics due to its portability, ease of use, and ability to monitor cerebral hemodynamics. BCIs traditionally rely on amplitude-based features for decoding motor activities, but the potential of phase connectivity features has been rarely explored. This study explores the feasibility of decoding contralateral finger movements using fNIRS signals with phase connectivity features. fNIRS data from five subjects performing motor tasks involving left- and right-hand finger tapping, along with a control condition, were analyzed. Preprocessing steps included noise reduction, conversion to oxyhemoglobin and deoxyhemoglobin concentrations, and extraction of statistical features and phase-based connectivity metrics, including phase-locking value (PLV) and magnitude-squared coherence (MSC), were employed. The classification of tapping conditions was conducted using various machine learning algorithms, including support vector machines (SVM), random forests (RF), k-nearest neighbors (k-NN), and artificial neural networks (ANNs). Our results demonstrate that combining statistical and phase-based connectivity features significantly enhances classification accuracy, with k-NN achieving the highest average performance of 96.1 ± 0.18%. These findings highlight the potential of fNIRS with phase connectivity features to decode fine motor movements, offering novel insights into neuroprosthetic applications and BCIs.