Today’s post is on the topic of motion contamination of fMRI data and how subject motion, motion correction and receive field contrast can conspire to create erroneous image-space signal and decrease temporal SNR (signal-to-noise ratio) of the EPI (echo planar image) time-series data used in fMRI (functional magentic resonance imaging). We have recently written a paper which reports on simulations of this RFC-MoCo (receive field contrast and motion correction) effect and we welcome feedback pertaining to the methods and results of the paper.

So what is the RFC-MoCo effect? The receive coil of an MRI scanner has an associated receive field which imparts a contrast to the image. This contrast, unlike intrinsic contrast (eg. BOLD, T1, T2 and proton density contrast), which is spatially fixed relative to the head, is instead spatially fixed relative to the scanner. Subject motion during the acquisition of the EPI time-series together with subsequently applied motion correction is equivalent (for perfect motion correction) to moving the head coil relative to the head. Since the receive field (or the composite sum-of-squares receive field of an receive array) is not spatially homogeneous then what results are temporal fluctuations in the time series data that are proportional to the local gradient of the receive field and the motion.

The paper examines the RFC-MoCo effect for two head receive coils: A 16-leg birdcage and a 12-channel receive array. The 12-channel data is reconstructed by means of the common sum-of-squares method. The simulations demonstrate that the RFC-MoCo effect is expected to be much greater for the array than for the birdcage. The paper also demonstrates that:

  1. The RFC-MoCo effect size is expected to be comparable to or greater than the BOLD (blood oxygen level dependent) effect of fMRI.
  2. The RFC-MoCo effect can introduce non-BOLD spatial-temporal correlations.
  3. The temporal SNR of the time-series can be reduced to such a degree that a receiver array – associated with increased tSNR for static objects – actually becomes a tSNR liability.

This should be of significance to fRMI researchers who study spatial-temporal correlations between regions of the brain – connectivity studies – in the resting state. It should, as well, be important to any fRMI researcher interested in acquiring data with better temporal SNR.

The paper mentions but does not go into any detail on potential fixes for the RCF-MoCo effect. Possible fixes to this problem include the application of a prescan normalization to remove the receive field contrast during the image reconstruction stage. Prescan normalization attempts to measure the receive field contrast by means of a single scan performed prior to acquisition of the EPI time-series data followed by removal of the contrast by a pointwise division of the contrasted image by the measured receive field. But presently the effectiveness of the prescan normalization method is unknown. So for now we have more concerns than fixes. Needless to say one thing is not in dispute – the need to keep your subjects still … very still.