Unsupervised physiological noise correction of functional magnetic resonance imaging data using phase and magnitude information (PREPAIR)

Abstract

Of the sources of noise affecting blood oxygen level-dependent functional magnetic resonance imaging (fMRI), respiration and cardiac fluctuations are responsible for the largest part of the variance, particularly at high and ultrahigh field. Existing approaches to removing physiological noise either use external recordings, which can be unwieldy and unreliable, or attempt to identify physiological noise from the magnitude fMRI data. Data-driven approaches are limited by sensitivity, temporal aliasing, and the need for user interaction. In the light of the sensitivity of the phase of the MR signal to local changes in the field stemming from physiological processes, we have developed an unsupervised physiological noise correction method using the information carried in the phase and the magnitude of echo-planar imaging data. Our technique, Physiological Regressor Estimation from Phase and mAgnItude, sub-tR (PREPAIR) derives time series signals sampled at the slice TR from both phase and magnitude images. It allows physiological noise to be captured without aliasing, and efficiently removes other sources of signal fluctuations not related to physiology, prior to regressor estimation. We demonstrate that the physiological signal time courses identified with PREPAIR agree well with those from external devices and retrieve challenging cardiac dynamics. The removal of physiological noise was as effective as that achieved with the most used approach based on external recordings, RETROICOR. In comparison with widely used recording-free physiological noise correction tools-PESTICA and FIX, both performed in unsupervised mode-PREPAIR removed significantly more respiratory and cardiac noise than PESTICA, and achieved a larger increase in temporal signal-to-noise-ratio at both 3 and 7 T.

Keywords: PREPAIR; fMRI; noise correction; phase data; physiological noise; unsupervised.

FIGURE 1

Steps in generating PREPAIR regressors (letters in brackets refer to steps described in Section 2.3). Raw magnitude and phase were preprocessed (for magnitude, with masking, for phase, with coil combination, unwrapping and masking). Magnitude and phase slices were averaged (a, b) then detrended and temporally reordered to yield slice TR sampled signals (c, d) from which respirators and cardiac time series were prefiltered (e). After deriving the corresponding physiological fundamental frequencies (f), the narrowed power spectra (g) underwent a variance test to choose between magnitude and phase physiological signals (h) to create the PREPAIR regressors (i)

FIGURE 2

Illustration of physiological noise identification in the slice TR sampled phase signal of subject S6, 3T_TR_700 (a), which is dominated by periodic slice effects (green arrows) sampled at f _1/TR, and slow fluctuations (blue square) comprising respiration (sampled at 1/fR) and cold head pump (sampled at 1/fpump). After removing the slice effects (b), physiological fluctuations were prefiltered from the phase signal (c) and unrelated physiological frequencies (sidebands SR and Spump) were removed to find the correct cardiac fundamental frequency f _C (d). Respiratory and cardiac signals were band‐pass filtered around f _R and f _C, respectively. Correlations improvement with the external signals are indicated for each step of the algorithm

FIGURE 3

Correlation of PESTICA (blue) and PREPAIR (red) with the externals signals across subjects for respiration and cardiac for the 3 T study (top) and 7 T study (bottom). Most mean correlations of PREPAIR are significantly (*p < .05; **p < .01; ***p < .001) above those of PESTICA

FIGURE 4

Spectrograms of uncorrected (left) versus corrected magnitude data. Frequencies (y axis) were truncated at 1.7 Hz and rows were scaled to the same power expressed in decibels (color scale). Horizontal black and red arrows indicate the first (and second when applicable) harmonic of the respiratory and cardiac noise, respectively, obtained with the external signals. PREPAIR showed effective noise removal in comparison with the other methods, even with unusual cardiac rates (top and bottom). Numbered vertical arrows show islets of remaining power after noise correction

FIGURE 5

Proportion of respiratory and cardiac noise removed by each method and for each protocol across subjects. Physiological noise correction performed with PREPAIR was as effective as that of RETROICOR‐EXT but more effective than FIX at 7 T only and PESTICA at both 3 and 7 T

FIGURE 6

Temporal signal‐to‐noise‐ratio (tSNR) gain map (S2, 3T_TR_700) in different anatomical regions (left panel). The level of improvement with PREPAIR was comparable with RETROICOR‐EXT and larger than with PESTICA. For all protocols and anatomical regions (right panel), tSNR gain with PREPAIR was significantly higher than PESTICA (*p < .05; **p < .01; ***p < .001). The level of tSNR improvement with PREPAIR in the insular and visual cortices is slightly weaker than RETROICOR‐EXT but comparable in the brainstem