Reduction of global interference in functional multidistance near-infrared spectroscopy using empirical mode decomposition and recursive least squares: a Monte Carlo study

Abstract

Functional near-infrared spectroscopy (fNIRS) is a sensitive technique that has the potential to detect haemodynamic changes during the performance of specific activation tasks. However, in real situations, fNIRS recordings are often corrupted by physiological phenomena, especially by cardiac contraction, breathing and blood pressure fluctuations, and these forms of interference can severely limit the utility of fNIRS. We present a novel fNIRS enhancement based on the multidistance fNIRS method with short-distance and long-distance optode pairs. With this method empirical mode decomposition (EMD) is applied to decompose the short-distance fNIRS measurement into a series of intrinsic mode functions (IMFs). By utilizing the weighting coefficients for the IMFs, we derive an estimate for global interference in the long-distance fNIRS measurements. We recover the evoked brain activity by minimizing least squares between the long-distance measurements and the estimated global interference. To accelerate the computation we adopt the recursive least squares (RLS) to decrease the computation complexity due to the matrix inversion. Monte Carlo simulations of photon propagation through a five-layered slab model of a human adult head were implemented to evaluate our methodology. The results demonstrate that the EMD-RLS method can effectively remove contamination from the evoked brain activity.