Motion artifact correction for resting-state neonatal functional near-infrared spectroscopy through adaptive estimation of physiological oscillation denoising

Abstract

Significance: Functional near-infrared spectroscopy (fNIRS) for resting-state neonatal brain function evaluation provides assistance for pediatricians in diagnosis and monitoring treatment outcomes. Artifact contamination is an important challenge in the application of fNIRS in the neonatal population. Aim: Our study aims to develop a correction algorithm that can effectively remove different types of artifacts from neonatal data. Approach: In the study, we estimate the recognition threshold based on the amplitude characteristics of the signal and artifacts. After artifact recognition, Spline and Gaussian replacements are used separately to correct the artifacts. Various correction method recovery effects on simulated artifact and actual neonatal data are compared using the Pearson correlation (R) and root mean square error (RMSE). Simulated data connectivity recovery is used to compare various method performances. Results: The neonatal resting-state data corrected by our method showed better agreement with results by visual recognition and correction, and significant improvements (R = 0.732 ± 0.155, RMSE = 0.536 ± 0.339; paired t-test, **p < 0.01). Moreover, the method showed a higher degree of recovery of connectivity in simulated data. Conclusions: The proposed algorithm corrects artifacts such as baseline shifts, spikes, and serial disturbances in neonatal fNIRS data quickly and more effectively. It can be used for preprocessing in clinical applications of neonatal fNIRS brain function detection.