Investigation of the impact of receive field sensitivity on motion corruption in 3D-EPI for fMRI
Nadine N Graedel1, Nadege Corbin1, Yael Balbastre1, Oliver Josephs1, and Martina F Callaghan1

1Wellcome Centre for Human Neuroimaging, UCL Institute of Neurology, London, United Kingdom


High temporal signal-to-noise ratio (tSNR) is crucial in fMRI to maximise functional sensitivity. The use of high-density receiver arrays can greatly improve tSNR and enables parallel imaging, a requirement for imaging with high spatial resolution while maintaining reasonable scan times. The 3D-EPI approach enables through plane acceleration but at the cost of increased motion sensitivity. Here we explore the impact of rapidly varying sensitivity fields on the degradation of tSNR in the presence of motion in the context of 3D-EPI.


Functional MRI (fMRI) is very motion sensitive because even a small amount of motion can generate signal changes larger than the BOLD signal of interest, reduce statistical significance of activation and increase the prevalence of false activations1,2. Achieving high spatial resolution while maintaining reasonable scan times necessitates the use of parallel imaging for which high density receiver arrays are typically used. The use of receive arrays with such highly-localised sensitivities complicates the problem of motion because the head moves relative to the stationary coils. This can cause additional motion artifacts, which may (partially) offset the benefit of high-density array coils3 and is thought to be particularly problematic for fMRI4. This is further accentuated when parallel imaging is used because the pre-calibrated coil sensitivity may no longer be valid when motion occurs5. The goal of this work is to assess the impact of different receiver profiles on the degradation of image quality, in particular tSNR, due to motion. This would ultimately help guide coil selection, particularly for cohorts prone to motion, such as patients with motor disorders or dementia.


We simulated time series data (50 volumes) in order to disentangle the interaction of the receive field sensitivity and motion in the context of 3D–EPI for fMRI applications. To ensure a realistic motion trajectory, the input motion for the simulations was measured on a subject using a bore-mounted prospective motion correction (PMC) camera (Kineticor). For each EPI readout motion was simulated by applying the measured translations/rotations to a complex reference image and subsequently extracting the relevant k-space plane (see Fig. 1). The impact of coil configurations was assessed by multiplying the image with measured complex coil sensitivities (estimated/extrapolated to the full FOV) from three different receivers (20/32/64ch). In the first simulation (“static coil sensitivities”) the image was multiplied by the coil sensitivities after the motion was applied, in the second the sensitivities were incorporated prior to the motion, modelling the coil sensitivity tracking the head (“dynamic coil sensitivities”). These two cases were compared to isolate the impact of differential coil sensitivity.

To isolate the impact of receive sensitivities empirically, data were collected on a healthy volunteer using a high-resolution whole-brain 3D-EPI protocol (64ch coil, acceleration factor R=2x2, CAIPI shift=1, resolution=1.5 mm isotropic, FOV=192x192x120 mm3, TR/TE/FA=70ms/35ms/17°, TRvol=2.8s). On the first run the subject was instructed to move. The subjects’ movements were measured using the PMC camera (no PMC performed). On a second run the participant was instructed to stay still, while the previously measured motion time course was played out by adjusting the gradients/RF chain (using functionality of the PMC system).


The drop in tSNR caused by motion-induced inconsistencies in the k-space data was large (~90% reduction in tSNR compared to time-series with thermal noise only). However, the further loss of tSNR due to static coil sensitivities (Fig. 2) was small (< 5 % difference in tSNR, static>dynamic.) The 20-channel coil had somewhat less tSNR degradation compared to the 32 and 64 channels, though the effect is small and warrants further investigation.

Consistent with this finding, our in-vivo experiment showed no substantial tSNR difference between the time series with subject motion and the time series where the same motion was introduced by the PMC system while the subject remained static (Fig. 3). In the latter experiment there was no relative motion between the participant and the coil.

The fact that the tSNR was largely unchanged (though greatly reduced relative to no-motion) confirmed the simulation result that the effect of sensitivity in 3D-EPI is a secondary/small effect, at least for the scale of motion tested.


The effect of participant motion relative to the static coil sensitivities was small in scale compared to the overall motion-induced tSNR drop. However, if perfect retrospective/prospective motion correction was achieved the effect of coil sensitivity would remain and may be sufficiently big to compete with BOLD-induced signal changes. In addition, the motion trajectories applied in this work were large. In future work we plan to explore how the relative effect of coil sensitivities scales with the amplitude/type of motion.

While there was good agreement between the simulations and experiment, the simulations did not take into account (1) parallel imaging, (2) the alteration of sensitivities due to position-specific coil loading or (3) the alteration of the susceptibility-induced field distributions.

Our simulations and preliminary data suggests that the relative impact of more spatially localised coil sensitivities on 3D-EPI data is less than might be expected, such that this would not offset the inherent benefits of tSNR/parallel imaging performance that these coils offer. However further simulations and in-vivo data are required to test this across different motion regimes.


The Wellcome Centre for Human Neuroimaging is supported by core funding from the Wellcome [203147/Z/16/Z].


1. K. J. Friston, S. Williams, R. Howard, R. S. Frackowiak, and R. Turner. “Movement-related e ects in fMRI time-series.” Magnetic Resonance in Medicine 35.3 (1996), pp. 346–355.

2. J. V. Hajnal, R. Myers, A. Oatridge, J. E. Schwieso, I. R. Young, and G. M. Bydder. “Artifacts due to stimulus correlated motion in functional imaging of the brain.” Magnetic Resonance in Medicine 31.3 (1994), pp. 283–291.

3. D. Sheltraw, and B. Inglis. “A Simulation of the Effects of Receive Field Contrast on Motion-Corrected EPI Time Series.” arXiv.org physics.med-ph:arXiv:1210.3633 (2012).

4. Zaitsev M, Akin B, LeVan P, Knowles BR. “Prospective motion correction in functional MRI. Neuroimage.” 2017;154:33–42.

5. Z. Faraji-Dana, F. Tam, J. J. Chen, and S. J. Graham. “Interactions between head motion and coil sensitivity in accelerated fMRI.” Journal of Neuroscience Methods 270 (2016), pp. 46-60.


Fig 1. Schematic of motion simulation pipelines: (a) “Static coil sensitivities”: The motion is applied prior to multiplication with the coil sensitivities. White noise is added to the real/imaginary channels of each coil. After generating the motion-corrupted k-space the samples corresponding to the current TR are extracted. This process is repeated until the final k-space is filled. Pink/green lines indicate example TRs and the corresponding k-space planes. All images and coil sensitivities are complex, only the magnitude is displayed. (b) “Dynamic coil sensitivities”: The sensitivities are multiplied before the motion is applied, in other words they track the head position.

Fig 2. Results of the motion simulations: (a) Simulated maps of tSNR for no motion (left), motion with static sensitivities (middle) and dynamic sensitivities (right) for 20/32/64 channel coils. (b) Bar plots of average tSNR (using a whole brain mask). (c) Bar plots of percent decrease of tSNR for the static coil sensitivities compared to the dynamic ones.

Fig 3. Comparison of tSNR maps for a 3D EPI time series corrupted by deliberate subject motion (left) and for the same motion played out by the gradients/RF chain while the subject remained still (right). Below the images the motion-related effects we are expecting to capture for the two cases are listed.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)