Disturbance elimination in near-infrared spectroscopy by correlated empirical mode decomposition

Abstract

Near-infrared spectroscopy (NIRS) is a noninvasive measurement method used to gain the information regarding brain activities through the scalp. Near-infrared light can penetrate several centimeters into the scalp tissue and indicate the oxygen concentration in the outer part of the cerebral cortex. Three major disturbances of NIRS are respiratory, the heartbeat, and Mayer waves. The nonlinear coupling between signals and noise can lead to the loss of meaningful information when linear noise-removal methods are applied. To resolve this problem, correlated empirical mode decomposition (CEMD), a new algorithm used to extract the common modes between two correlated signals, was adopted to separate the signal and disturbance. Public NIRS data were used as a pure input signal, and respiratory and electrocardiogram signals were added to it. The correlated intrinsic mode functions associated with the known disturbances were eliminated through the CEMD screening process. The spectrum results indicated that the energy of the additional signals can be eliminated. The CEMD method can eliminate the disturbances and thus reduce their effects on signal analyses.