Improvements in data acquisition and artefact correction for diffusion MRI

Diffusion magnetic resonance imaging (MRI) and in particular diffusion tensor imaging (DTI) is widely used in research and clinical applications, but still suffers from substantial artefacts. Remedies for these artefacts can be grouped in two different categories: 1) modification of the data acquisition (pulse sequence programming) and 2) post-processing correction of the acquired data. Even in the absence of gross artefacts, improvements to the acquisition efficiency are possible. Our group works on methods in all of these areas.



1) Fat suppression with slice select gradient reversal (SSGR)


In many different MR imaging sequences signal from lipids interferes with the signal of interest from water. Measures must be taken to eliminate this unwanted signal contribution. This however takes extra time, increases load on the hardware and increases the exposure of the subject to radiofrequency pulses. We developed a method [1], which eliminates the signal from lipid without increasing scan time, hardware usage or RF deposition when a twice-refocused spin echo sequence is used for diffusion measurements [2].


Figure 1 – SSGR results


Images of a composite phantom including a layer of lipid. The middle image indicates the artefact. On the left the fat suppression method increases imaging time, system load and RF deposition. The image on the right was collected with the SSGR method showing an efficient and complete removal of the artefactual signal.



DTI artefacts result from the application of strong diffusion sensitizing gradients. Eddy currents (EC), gradient nonuniformity and miscalibration, can cause local perturbation fields (LPFs) or lead to vibration of the patient table. These artefacts can increase the noise in diffusion tensor estimates or even bias quantitative DTI/diffusion MRI indices and thus counteract future developments.

2) Correcting for the effect of local peturbation fields

The nonuniformity and miscalibration of gradients and other LPFs like the ECs can lead to a local mismatch between the effective and the expected diffusion gradients, resulting in a spatially varying error in the diffusion weighting B matrix and diffusion tensor estimation (see Fig. 2).



Figure 2: Example for local mismatch between expected (blue) and effectively applied (red) diffusion gradient. Local perturbation fields due to eddy currents (EC) lead to a spatially-dependent error in the B matrix that affect the FA contrast. Note that the depicted EC effects are exaggerated.

One underlying assumption of diffusion imaging is that one should measure similar values for the apparent diffusion coefficient (ADC) from a water-filled phantom a) in all voxels and b) regardless of the direction of the diffusion encoding gradients. As a result, a water phantom can be used to measure the error in the ADC (Fig. 3) [3–5] and estimate the LPFs [3], [5]. The estimated LPFs can be either used to correct the applied diffusion gradients directly in the diffusion sequence (Fig. 4) [3] or to retrospectively correct for the error in the B matrix (Fig. 3) [5]. An SPM toolbox that uses the methods described in [5] to estimate and correct the effects of LPFs is under construction.


Figure 3 – Gradient Non-linearity

Two different diffusion-encoding directions before (top) and after (bottom) correction. This is an example of post-processing the acquired data to correct the b-value used for measuring the ADC.

Figure 4 – Gradient miscalibration

If the magnetic gradients are not calibrated identically, the diffusion encoding b-value can be axis dependent (left column). Middle column shows the variability of ADC depending along 60 isotropically distributed diffusion directions befor (top) and after (bottom) correction. From the rightmost column it is clear that the variability of ADC is systematic and due to the gradient axis (X, Y and Z) used before (top) but not after (bottom)

3) Retrospective eddy current and motion correction


ECs lead to image distortions (Fig. 5), which differ for different diffusion gradients [6]. Using the non-diffusion weighted image as a reference, EC-induced image distortions and rigid-body subject motion can be corrected, retrospectively (see e.g. the ECMOCO toolbox).


Figure 5:
Eddy currents can lead to image distortions of the high-b-value diffusion-weighted images that depend on the amplitude, duration, and orientation of the applied diffusion gradients (second and third row). A less distorted low-b-value image (first row) can be used as reference to correct for affine EC distortions.

4) Correction of vibration artefacts in DTI using phase encoding reversal


Vibration of the patient table that are induced by the strong diffusion gradients in DTI can lead to an echo shift in k-space and consequential signal-loss in diffusion weighted images (Fig. 3). We showed that asymmetric k-space coverage in widely used Partial Fourier acquisitions results in locally differing signal loss in images acquired with reversed phase encoding direction (blip-up/blip-down images) [7]. Based on this theory and using a similar well established approach for reducing susceptibility-induced signal loss due to echo shifts in gradient echo echo-planar imaging [8], [9], a correction of vibration artefacts in DTI has been developed. For this correction, an SPM toolbox is under construction.



Figure 6: Example for signal-loss (left image) due to patient table vibration caused by strong diffusion gradients (arrow). The signal-loss leads to arbitrarily high FA values at the parietal lobe (middle image) and apparent left-right connection, which is visible on the FA image that is overlaid with the principal diffusion direction (right image).

Primary contact

Siawoosh Mohammadi (siawoosh.mohammadi «at»


[1]        Z. Nagy and N. Weiskopf, “Efficient fat suppression by slice-selection gradient reversal in twice-refocused diffusion encoding,” Magnetic Resonance in Medicine, vol. 60, no. 5, pp. 1256-1260, Nov. 2008, doi: 10.1002/mrm.21746.

[2]        T. G. Reese, O. Heid, R. M. Weisskoff, and V. J. Wedeen, “Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo,” Magnetic Resonance in Medicine, vol. 49, no. 1, pp. 177-182, Jan. 2003.

[3]        Z. Nagy, N. Weiskopf, D. C. Alexander, and R. Deichmann, “A method for improving the performance of gradient systems for diffusion-weighted MRI,” Magnetic Resonance in Medicine, vol. 58, no. 4, pp. 763-768, Oct. 2007, doi: 10.1002/mrm.21379.

[4]        Z. Nagy, D. Alexander, and N. Weiskopf, “Measuring and correcting errors that occur in diffusion weighted images due to non-ideal gradient linearity,” in Proc Intl Soc Magn Reson Med. 2009;17:1377., 2009.

[5]        S. Mohammadi et al., “The effect of local perturbation fields on human DTI: Characterisation, measurement and correction,” NeuroImage, vol. 60, no. 1, pp. 562-570, Mar. 2012, doi: 10.1016/j.neuroimage.2011.12.009.

[6]        S. Mohammadi, H. E. Möller, H. Kugel, D. K. Müller, and M. Deppe, “Correcting eddy current and motion effects by affine whole-brain registrations: evaluation of three-dimensional distortions and comparison with slicewise correction,” Magnetic Resonance in Medicine, vol. 64, no. 4, pp. 1047-1056, Oct. 2010, 10.1002/mrm.22501.

[7]        S. Mohammadi, Z. Nagy, C. Hutton, O. Josephs, and N. Weiskopf, “Correction of vibration artifacts in DTI using phaseencoding reversal (COVIPER),” Magnetic Resonance in Medicine, doi: 10.1002/mrm.23308.

[8]        N. Weiskopf, U. Klose, N. Birbaumer, and K. Mathiak, “Single-shot compensation of image distortions and BOLD contrast optimization using multi-echo EPI for real-time fMRI,” NeuroImage, vol. 24, no. 4, pp. 1068-1079, Feb. 2005, doi: 10.1016/j.neuroimage.2004.10.012.

[9]        R. Deichmann, O. Josephs, C. Hutton, D. R. Corfield, and R. Turner, “Compensation of susceptibility-induced BOLD sensitivity losses in echo-planar fMRI imaging,” NeuroImage, vol. 15, no. 1, pp. 120-135, Jan. 2002, doi: 10.1006/nimg.2001.0985.

back to physics pages