A Systematic Comparison of Motion Artifact Correction Techniques for Functional Near-Infrared Spectroscopy

Abstract

Near-infrared spectroscopy (NIRS) is susceptible to signal artifacts caused by relative motion between NIRS optical fibers and the scalp. These artifacts can be very damaging to the utility of functional NIRS, particularly in challenging subject groups where motion can be unavoidable. A number of approaches to the removal of motion artifacts from NIRS data have been suggested. In this paper we systematically compare the utility of a variety of published NIRS motion correction techniques using a simulated functional activation signal added to 20 real NIRS datasets which contain motion artifacts. Principle component analysis, spline interpolation, wavelet analysis, and Kalman filtering approaches are compared to one another and to standard approaches using the accuracy of the recovered, simulated hemodynamic response function (HRF). Each of the four motion correction techniques we tested yields a significant reduction in the mean-squared error (MSE) and significant increase in the contrast-to-noise ratio (CNR) of the recovered HRF when compared to no correction and compared to a process of rejecting motion-contaminated trials. Spline interpolation produces the largest average reduction in MSE (55%) while wavelet analysis produces the highest average increase in CNR (39%). On the basis of this analysis, we recommend the routine application of motion correction techniques (particularly spline interpolation or wavelet analysis) to minimize the impact of motion artifacts on functional NIRS data.

Measuring tissue hemodynamics and oxygenation by continuous-wave functional near-infrared spectroscopy—how robust are the different calculation methods against movement artifacts?

“…Therefore, application of specially designed signal processing algorithms will further reduce MAs, e.g. (Barker et al 2013, Cooper et al 2012, Izzetoglu et al 2005, Molavi and Dumont 2011, Robertson et al 2010, Sato et al 2006, Scholkmann et al 2010.…”

“…Barker et al 2013, Cooper et al 2012, Izzetoglu et al 2005, Molavi and Dumont 2011, Robertson et al 2010, Sato et al 2006, Scholkmann et al 2010. Besides the signal post-processing, the actual type of measurement methods used determines the impact of the MAs on the resulting signals (Yamada et al 2009).…”

A New Approach for Automatic Removal of Movement Artifacts in Near-Infrared Spectroscopy Time Series by Means of Acceleration Data

“…MARA has been successfully applied in several NIRS studies so far [5][6][7][8][9]12,13,16,[35][36][37][38][39]. Its performance was recently evaluated positively in a study comparing different MA correction techniques for NIRS [40]. However, the limitations of MARA were highlighted in a comparison of five different MA reduction algorithms [41].…”

Correction of motion artifacts and serial correlations for real-time functional near-infrared spectroscopy

“…In addition, real-time correction methods need to be forward directed, meaning that they need to offer corrections using only past history data points. In contrast, many of the motion correction methods, such as wavelet or spline interpolation models, 26,28 use data information from both before and after the artifact for correction and thus only offer retrospective corrections.…”

“…In particular, the two major sources of confounding noise that affect the analysis and interpretation of fNIRS signals are serially correlated errors due to systemic physiology, such as cardiac, respiratory, and low-frequency Mayer waves (related to blood pressure regulation), and motion artifacts due to the movement or slippage of the head cap. While several approaches to offline correction of motion [25][26][27][28] and physiological noise [29][30][31] have been proposed, for real-time imaging, these corrections must be both automated and quickly implemented to keep up with the high sample rates of fNIRS systems. In addition, real-time correction methods need to be forward directed, meaning that they need to offer corrections using only past history data points.…”