Nino Kobalia^{1}, Farida Grinberg^{1,2}, Ezequiel Farrher^{1}, Xiang Gao^{1}, and N. Jon Shah^{1,2}

^{1}Institute of Neuroscience and Medicine - 4, Forschungszentrum Jülich GmbH, Jülich, Germany, ^{2}Department of Neurology, Faculty of Medicine, JARA, RWTH Aachen University, Aachen, Germany

### Synopsis

**The knowledge of intra- and inter-subject variability of
diffusion kurtosis imaging metrics plays an important role in the
interpretation of the results in clinical trials. However, it has not been
sufficiently studied thus far. The purpose of this work is to investigate
between-session variability of a single subject with N-repeated measurements
with an identical experimental protocol, and thus to provide the baseline for
comparison with phantom measurements and inter-subject in vivo variability. We
quantified variability in terms of the coefficient of variation and studied how
its value varies between various diffusion tensor and kurtosis metrics
estimated in twenty anatomical regions.**### Target
audience

The target audience of this abstract is researchers
investigating non-Gaussian water diffusion properties in biological tissues.

### Purpose

Diffusion kurtosis imaging (DKI)

^{1} is a method used to
quantify the non-Gaussian diffusion of water in biological tissues. It provides
complementary information on white matter microstructure to that provided by
the conventional diffusion tensor imaging (DTI) technique. However, in
comparison to DTI, variability of DKI metrics is rarely studied. Previously
variability was investigated for two DKI parameters

^{2,3}, mean kurtosis (MK) and
radial kurtosis (RK), in three major white matter structures, such as superior
cingulum bundle (CG), medial motor corticospinal tract (CST), and mid-sagittal
corpus callosum (CC). The aim of this work is to perform an investigation of
single-subject between-session variability of all typical DTI/DKI parameters
and its dependence on the anatomical region-of-interest. The data are to be
compared with between-session variability of an anisotropic synthetic physical
phantom additionally, and should also provide the baseline for a comparison
with inter-subject variability.

### Materials and methods

In vivo brain DKI data sets in healthy subjects (N=3) and a
physical synthetic fibre phantom

^{4} were acquired on a 3T Siemens MAGENETOM Tim
Trio scanner using the following protocol: four b-values, b = 0, 700, 1000,
2500 s/mm

^{2}; 150 diffusion gradient directions; TE/TR=109ms/9000ms; voxel size =
2x2x2 mm

^{3}. The measurements were repeated in 10 different sessions for each
subject and the phantom. Eddy current as well as motion distortions were
corrected using the tool “EDDY”

^{5}. The corresponding rotation matrix was applied
to the diffusion gradient directions. Bias due to background noise was
corrected using the power-images method

^{6}. Gibbs ringing artefacts were
corrected using the Total Variation method

^{7, 8}. Finally, DKI parameters were
evaluated using a weighted-linear least-squares approach available in the
ExploreDTI toolbox. The diagonal elements of the kurtosis tensor were
constrained to be positive. Variability
was quantified using the coefficient of variation (CV), that represents the
relative standard deviation. CV was estimated based on 10 session measurements
for a total of 20 white matter tracts provided by the John Hopkins University
(JHU) atlas

^{9}. The investigated WM structures comprised the left and right
regions of 7 major association fibres, such as cingulum (gyrus) (Cg) and
cingulum (hippocampus) (Ch), superior longitudinal fasciculus (SLF), SLF
(temp), inferior longitudinal fasciculus (ILF), inferior fronto-occipital
fasciculus (IFOF), uncinate fasciculus (UF), the left and right regions of 2
projection fibres, anterior thalamic radiation (ATR) and corticospinal tract
(CST), and 2 commissural fibres, forceps major (F_major) and forceps minor
(F_minor). Left and right regions of the same fibre will be denoted by
subscripts “L” or “R”.

### Results
and discussions

Figures 1 and 2 show the CV values for DTI (Fig. 1) and DKI
(Fig. 2) metrics for various fibres. One can see that CV strongly differs for
various metrics and between various regions. Moreover, the CV values for DKI
metrics tend to be larger than on DTI alone metrics, by approximately a factor
of 1.5 – 2.0. The difference of CV for the DTI metrics for various fibres is
especially high for UF, SLF (temp) and CH fibres. CV in FA metrics on ATR_L,
SLF_L and ILF_R fibers is in the range between 0.015-0.025, different from
other fibers where CV is below 0.008. MD range on UF_L and SLF (temp) fibers
varies between 0.014-0.019, but for the rest of the fibers they are below
0.012. CV values of AD for CH, UF, and SLF (temp)_L is between 0.015-0.025, but
for the rest of fibres they are below 0.015.CV values for RD for UF and SLF
(temp) range is between 0.017-0.032, where as for the rest of fibers they are
below 0.015. DKI shows significantly different behaviour from DTI. In
particular, CV of KA is high for all fibres (0.015-0.048), but very low
(0.01-0.02) on ATR, CST, CG, and CH_R. CV of MK stays in the same range between
(0.07-0.034) that is similar to CV on all fibres for MD. CV for AK varies
between (0.01-0.05) and for RK range for CV is 0.012-0.046, where the highest
variability is shown on CH_L (0.032), F_minor (0.037) and SLF (temp)_L (0.047).
We shall discuss our findings in the context of various reasons contributing to
DTI/DKI parameter variability, in particular, those that are general for the
method and those that are specific for in vivo brain.

### Conclusion

We demonstrate that all DTI/DKI parameters show
heterogeneous single-subject, between-session variability in a number of
anatomical regions. Our results provide the baseline for a comparison with
inter-subject variability and should be useful for better interpretation of
clinical studies subject to both intra- and inter-subject variability.

### Acknowledgements

No acknowledgement found.### References

[1]
Jensen J.H, Helpern J.A, Ramani A, et al. Diffusional kurtosis imaging: The quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magnetic Resonance in Medicine. 2005; 53: 1432-1440.

[2] Szczepankiewicz F, Lätt J, Wirestam R, et al. Variability in diffusion kurtosis
imaging: Impact on study design, statistical power and interpretation. NeuroImage. 2013; 76: 145-154.

[3] André E.D, Grinberg F, Farrher E, et al. Influence of Noise Correction on Intra- and
Inter-Subject Variability of Quantitative Metrics in Diffusion Kurtosis
Imaging. Plos one. 2014; 9: e94531.

[4] Farrher E, Kaffanke J, A. Celik A, et al. Novel multisection design of
anisotropic diffusion phantoms. Magnetic Resonance Imaging. 2012; 30: 518-526.

[5] Woolrich M.W, Jbabdi S, Patenaude B, et al. Bayesian analysis of neuroimaging
data in FSL. NeuroImage. 2009; 45:173-86.

[6] Miller A.J, Joseph P.M. The use of power images to
perform quantitative analysis on low SNR MR images. Magnetic Resonance Imaging. 1993;
11: 1051-1056.

[7] Leemans A, Jones D.K. The B-Matrix Must Be Rotated When
Correcting for Subject Motion in DTI Data. Magnetic Resonance in Medicine. 2009;
61:1336–1349.

[8] Leemans A, Jeurissen B, Sijbers J, et al. ExploreDTI: a graphical toolbox for
processing, analyzing, and visualizing diffusion MR data. Hawaii, USA: ISMRM; 2009.

[9] Mori S, Oishi K, Jiang H, et al. Stereotaxic White Matter Atlas Based on
Diffusion Tensor Imaging in an ICBM Template. Neuroimage. 2008; 40: 570–582.