Physiological Noise Extraction in fMRI Data Using Empirical Mode Decomposition

Introduction Physiological noise caused by ecgand/or breathing related pulsatility is known to have substantial influence on the BOLD signal [1,2]. Such effects may introduce temporal correlations that are unrelated to neuronal processes in a resting-state network analysis [3]. Various noise removal techniques have been proposed, including image-based estimation schemes [4,5] and methods that use external ECG/respiration measurements as regression models such as RETROICOR [6]. Empirical mode decomposition (EMD) has been shown to have good performance on analyzing non-linear, non-stationary data [7]. It separates signals into intrinsic mode functions (IMFs) that are more likely to yield real physical meaning than simple filtering in the Fourier domain. In this study we implemented EMD on resting-state fMRI time-series and extracted cardiac components, then compared it’s time-frequency curve with ECG recordings. Fixed band-pass filtering and RETROICOR results were also obtained for comparison.