Image Artifacts & Processing Pipelines, Part I
Rita G. Nunes1 and Jelle Veraart2,3

1ISR-Lisboa/LARSyS and Department of Bioengineering, Instituto Superior Técnico – Universidade de Lisboa, Lisbon, Portugal, 2Champalimaud Centre for the Unknown, Lisbon, Portugal, 3imec - Vision Lab, University of Antwerp, Antwerp, Belgium

Synopsis

Diffusion-weighted images (DWI) are corrupted by noise and various imaging artifacts such as Gibbs ringing, EPI and eddy current distortions, motion and other physiological artifacts. The correction of those artifacts is of utmost importance to improve the qualitative, quantitative and statistical inspection of the diffusion data. Here we will give an overview of the major image artifacts, explain how they might confound the DWI analysis, and how they can be corrected for or at least minimized at source or using image processing


Introduction

A side-effect of making Diffusion-Weighted images (DWI) able to detect the microscopic motion of water molecules is that they become extremely sensitive to subject motion occuring during the diffusion preparation module; this results in spatially-varying image phase patterns1. To avoid data inconsistencies, single-shot Echo Planar Images (EPI) are typically used to cover k-space after one single excitation2. Its high rate of data acquisition is also advantageous, enabling to acquire several images with varying diffusion-weighting and/or diffusion directions within reasonable exam times.

Artifacts due to Static Field Inhomogeneity

Unfortunately, EPI do suffer from other type of image artifacts. Its long readout window makes the images very vulnerable to static B0 field inhomogeneities resulting in geometric distortions, more pronounced along the slower phase-encode (PE) direction3. It is important to correct for this effect to accurately match the diffusion information to its anatomical location. One option is to acquire DWI with opposing PE directions so that regions which appear stretched in one image will look compressed in the other and vice-versa. This approach enables to estimate the local B0 field and to generate undistorted DWI4.

Chemical Shift Artifacts

EPI also suffers from chemical shift artifacts, with incorrect spatial encoding of signal originating from fat, which appears shifted by several pixels. Given that fat has a lower diffusion coefficient compared to water, fat signal is less attenuated in DWI. Effective fat suppression is particularly relevant when applying DWI outside the brain. This can be achieved with a slight modification to the diffusion preparation module, often based on spin-echoes, which consists of inverting the polarity of one of the slice-selection gradients5.

Geometric Distortions due to Eddy Current Fields and Motion-related Outliers

Another type of image artifact arises from the diffusion preparation module, due to the need to apply large amplitude gradients during long periods. Every time the gradients are switched on/off, eddy current fields (ECF) are produced which may persist during the EPI readout period, leading to spatial encoding errors. As described early on, the type of geometric distortion depends on the direction of the ECF6, and hence on the direction of the diffusion gradients. Given that estimation of diffusion parameters is carried out on a pixel-by-pixel basis, it is crucial to ensure spatial consistency across diffusion volumes. A standard approach is to null the dominant ECF component by using a modified diffusion preparation module (doubly-refocused), introducing a second refocusing pulse and adjusting the duration of each of the four diffusion gradients7. ECF components with other time constants may nonetheless still affect the images, with the distortions becoming increasingly more severe for higher levels of diffusion-weighting. Post-processing correction is often performed using a non-parametric motion and distortion estimation which can also incorporate rejection of outliers8. These can result from bulk subject motion, leading to signal loss in whole slices9, or have a more localised nature if arising for example from pulsatile brain motion10.

Signal-to-Noise ratio, From Thermal noise to the Rician bias

Thermal noise is an important source of MR signal distortion, especially in DWI where the sensitization to the diffusion process is characterized by strong signal decay, typically due to a combination of strong diffusion weighting and typically long echo times. The noise in the complex-valued MR signals is additive and normally distributed11, but can be spatially varying due to the use of parallel imaging and image reconstruction techniques for accelerated imaging12. The low SNR limits the precision and, albeit less intuitive, the accuracy of diffusion parameter estimators. The calculation of the signal’s magnitude to avoid signal drop-outs in the diffusion-weighted images due to motion-induced random phase shifts, will alter the noise characteristics of the DWI data, especially in case of low SNR13. These alterations in the noise distribution bias conventional diffusion parameter estimators, potentially interfering with the interpretation and analysis of diffusion parameter maps. Notorious examples are the squashing of the diffusion tensor “peanut”14 or the apparent non-Gaussian signal decay15. We will discuss how this so-called “Rician bias” affects DWI analysis and modeling, and how we can resolve this phenomenon.

Acknowledgements

JV is a Postdoctoral Fellow of the Research Foundation - Flanders (FWO; grant number 12S1615N)RGN has received funding from the POR Lisboa 2020 program (grant number LISBOA-01-0145-FEDER-029686).

References

[1] Anderson, A. and Gore, J., 1994, Analysis and correction of motion artifacts in diffusion weighted imaging. Magnetic resonance in medicine, 32(3), pp.379-87

[2] Turner, R., Le Bihan, D., Maier, J., Vavrek, R., Hedges, L., and Pekar, J., 1990, Echo-planar imaging of intravoxel incoherent motion. Radiology, 177(2), pp.407-414

[3] Jezzard, P., and Balaban, R., 1995, Correction for geometric distortion in echo planar images from B0 field variations. Magnetic Resonance in Medicine, 34(1), pp.65-73

[4] Andersson, J., Skare, S., and Ashburner, J., 2003, How to correct susceptibility distortions in spin-echo echo-planar images: Application to diffusion tensor imaging. NeuroImage, 20(2), pp.870-888

[5] Nagy, Z. and Weiskopf, N., 2008. Efficient fat suppression by slice-selection gradient reversal in twice-refocused diffusion encoding. Magnetic resonance in medicine, 60(5), pp.1256-60.

[6] Jezzard, P., Barnett, A.S., Pierpaoli, C., 1998. Characterization of and correction for eddy current artifacts in echo planar diffusion imaging. Magnetic resonance in medicine, 39(5), pp.801-12.

[7] Reese, T.G., Heid, O., Weisskoff, R.M. and Wedeen, V.J., 2003. Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo. Magnetic resonance in medicine, 49(1), pp.177-82

[8] Andersson, J., Graham, M., Zsoldos, E., and Sotiropoulos, S., 2016. Incorporating outlier detection and replacement into a non-parametric framework for movement and distortion correction of diffusion MR images. NeuroImage, 141 pp.556-572

[9] Wedeen, V.J., Weisskoff, R.M. and Poncelet, B.P., 1994. MRI signal void due to in-plane motion is all-or-none. Magnetic resonance in medicine, 32(1) pp.116-20.

[10] Wirestam, R., Greitz, D., Thomsen, C., Brockstedt, S., Olsson, M.B., Stahlberg, F., 1996. Theoretical and experimental evaluation of phase-dispersion effects caused by brain motion in diffusion and perfusion MR imaging. Journal of magnetic resonance imaging, 6 pp.348–55.

[11] Gudbjartsson, H. and Patz, S., 1995. The Rician distribution of noisy MRI data. Magnetic resonance in medicine, 34(6), pp.910-914.

[12] Dietrich, O., Raya, J.G., Reeder, S.B., Ingrisch, M., Reiser, M.F. and Schoenberg, S.O., 2008. Influence of multichannel combination, parallel imaging and other reconstruction techniques on MRI noise characteristics. Magnetic resonance imaging, 26(6), pp.754-762.

[13] Henkelman, R.M., 1985. Measurement of signal intensities in the presence of noise in MR images. Medical physics, 12(2), pp.232-233.

[14] Jones, D.K. and Basser, P.J., 2004. “Squashing peanuts and smashing pumpkins”: how noise distorts diffusion‐weighted MR data. Magnetic Resonance in Medicine, 52(5), pp.979-993.

[15] Veraart, J., Van Hecke, W. and Sijbers, J., 2011. Constrained maximum likelihood estimation of the diffusion kurtosis tensor using a Rician noise model. Magnetic Resonance in Medicine, 66(3), pp.678-686.

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