Same brain, different look? – A scanner and preprocessing pipeline comparison for diffusion imaging

Poster No:


Submission Type:

Abstract Submission 


Ronja Thieleking1, Rui Zhang1, Alfred Anwander1, Arno Villringer1, A. Veronica Witte1


1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany


Large population and longitudinal studies need to deal with data collection at different imaging sites or technical scanner changes in the course of the study. This can result in systematic errors biasing data analysis and interpretation. Similar obstructions may arise from the choice of preprocessing pipelines which can have a severe influence on anatomical and structural measures. This is specifically relevant for diffusion weighted imaging (DWI) where complex imaging artefacts augment the need to apply correction techniques during preprocessing. Recently, a preprocessing step has been introduced¹ that effectively reduces Gibbs-ringing (GR) artefacts, a common oscillation artefact in MRI which might cause physically implausible negative diffusivity and wrong fractional anisotropy (FA) values . Frequently used outcome measures from DWI include FA values providing information about white matter (WM) coherence, yet how different scanners and processing pipelines affect this outcome measure is largely unknown.
We aim at systematically assessing the impact of data collection at different imaging sites of the same participants and of data processing with varying pipelines on DWI data.


We collected DWI scans of 121 healthy participants (60f, 19-54 years) on two different 3T Siemens Magnetom scanners (Verio and Skyra, respectively) (b=1000, 60 dir, 7 b0, 1.7mm iso, GRAPPA 2, bipolar, TE 100 ms, MB 1, raw filter, CMRR sequence²). The preprocessing pipelines were modified by adding denoising (MRTRIX⁶) and removal of Gibbs-ringing ("unringing" ¹) followed by outlier replacement, motion correction and tensor fitting. The entire processing pipeline is depicted in figure 1.
Supporting Image: figures_prepro_1000.png
   ·Overview over entire processing pipeline


We obtained mean FA values of the whole brain WM skeleton⁵ and within 8 regions of interest (ROI) from the JHU DTI-based WM atlas (1mm)³, namely 4 in the corpus callosum (CC), the superior longitudinal fasciculus as well as the uncinate fasciculus (L and R respectively). ROIs were selected based on their different white matter characteristics. The comparison between scanners revealed a difference of about 0.9% in mean FA on the whole brain skeleton (see figure 2a). Nevertheless, scanner differences were not consistent throughout the whole brain: They varied in sign and magnitude from ROI to ROI–up to 4.3% in the CC genu (see figure 2b).
GR artefacts are strongest in the b0 images but are immensely reduced after "unringing" with the Kellner tool compared to raw, only denoised or low-pass filtered data. In addition, the amount of voxels with implausible FA values > 1 is significantly lowered by "unringing" ¹. Through tract-based spatial statistics (TBSS)⁵ we observe the expected decrease of FA values as a consequence of ageing. This age effect is preserved in all preprocessing pipelines.
Supporting Image: Figure2.jpg
   ·Scanner und preprocessing pipeline differences in mean FA value of the white matter skeleton of a) the whole brain and b) four ROIs with different white matter characteristics.


The scanner difference of ~1% of the mean FA value on the skeleton is an alarmingly large effect as the age effect on FA comprises only a decrease of 0.14% per year (estimated based on additional analysis of data from the LIFE Adult Study, n=1255⁷). The regional variance in differences across the 8 ROIs revealed that a whole brain correction factor–as suggested by Pohl et al.⁴ to account for the systematic error introduced by scanner differences–is not applicable. However, correction factors for ROI-specific hypotheses might be possible. As the mean FA values show large differences between scanners even of the same manufacturer–with an effect size exceeding the age effect by more than 10 times–large multi-centre studies should account for the affected comparability of DWI data.
The improvements by reducing GR artefacts points to the need of using the "unringing" tool as a standard step in processing DWI images. Importantly, removing the artefact does not obscure physiological effects like ageing.

Imaging Methods:

Diffusion MRI 1



Modeling and Analysis Methods:

Diffusion MRI Modeling and Analysis 2
Motion Correction and Preprocessing


Data analysis
White Matter

1|2Indicates the priority used for review

My abstract is being submitted as a Software Demonstration.


Please indicate below if your study was a "resting state" or "task-activation” study.


Healthy subjects only or patients (note that patient studies may also involve healthy subjects):

Healthy subjects

Was any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.


Was any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.

Not applicable

Please indicate which methods were used in your research:

Structural MRI
Diffusion MRI

For human MRI, what field strength scanner do you use?


Which processing packages did you use for your study?

Other, Please list  -   MRTRIX

Provide references using author date format

1. Kellner, E. (2016), 'Gibbs-ringing artifact removal based on local subvoxel-shifts', Magnetic Resonance in Medicine, vol. 76, no. 5, pp. 1574-1581.
2. Moeller, S. (2010), 'Multiband multislice GE-EPI at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fMRI', Magnetic Resonance in Medicine, vol. 63, no. 5, pp. 1144-1153.
3. Mori, S. (2005), 'MRI Atlas of the Human White Matter', 1st ed., Elsevier, Amsterdam, The Netherlands.
4. Pohl, K.M, 'Harmonizing DTI measurements across scanners to examine the development of white matter microstructure in 803 adolescents of the NCANDA study', NeuroImage, vol. 130, pp. 194-213.
5. Smith, S.M. (2006), 'Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data', NeuroImage. vol. 31, no. 4, pp. 1487-1505.
6. Veraart, J. (2016), 'Diffusion MRI noise mapping using random matrix theory', Magnetic Resonance in Medicine, vol. 76, no. 5, pp. 1582-1593.
7. Zhang, R. (2018), 'White Matter Microstructural Variability Mediates the Relation between Obesity and Cognition in Healthy Adults', NeuroImage, vol. 172, pp. 239-249.