0411
Advantages of tractography for the evaluation of intramuscular variances in human thigh muscles1Neurology, BG-University Hospital Bergmannsheil, Bochum, Germany, 2Radiology, University Medical Center Utrecht, Utrecht, Netherlands
Synopsis
Muscle diffusion tensor imaging can indirectly provide information about muscular microstructure and architecture, which plays an increasing role in the evaluation of neuromuscular disease progression and treatment monitoring. The separation of different muscles is essential to evaluate intermuscular differences and variances. Here we have compared three methods to assess diffusion metrics of thigh muscles and showed, that tractography shows less variance in diffusion metrics than parameter maps. For the observed parameters FA, MD, RD and λ1 we found significant main effects of muscle, when using tractography, which was not found with manual annotation using ROI assessment of the parameter maps.
Introduction
Magnetic resonance imaging methods
play an increasing role in the assessment of muscle tissue particularly in the evaluation
of neuromuscular disease progression and treatment monitoring. Muscle diffusion
tensor imaging (mDTI) can indirectly provide information about muscular
microstructure and architecture. However, due to intermuscular variations of
the diffusion parameters the correct annotation of different muscles is
essential 1.
The purpose of this study was to
compare three different methods to evaluate diffusion metrics (fractional
anisotropy - FA, mean diffusivity - MD, radial diffusivity -RD) with respect to
their ability to indicate intramuscular differences.Methods
Data from thigh muscles of 30 healthy volunteers were acquired on a Philips Achieva 3T system. The protocol included a T1w scan for an anatomical reference and a spin-echo EPI to acquire DWIs with 17 gradient orientations and three non-weighted images. The entire thigh region was split into three fields of view to avoid shimming artifacts. Data were denoised and corrected for motion and eddy current distortions. The stacks were joined according to Schlaffke et al., 2017 and Froeling et al., 2015. REKINDLE2 was used to detect and remove outliers in combination with the WLLS estimation approach3 to compute the diffusion tensor 4,5. Next, parameter maps for FA, MD, RD and λ1 were calculated. Deterministic tractography for both legs was performed using a 3 x 3 x 3 mm3 seed resolution. The minimum FA value to select seed points / allow tractography was chosen to be 0.1. The maximal FA to allow tractography was set to 0.6. Tracking was stopped if the angle change exceeded 15° per 1.5 mm propagation step. Six thigh muscles (biceps femoris, semimembranosus, semitendinosus, rectus femoris, vastus lateralis and vastus medialis) were manually segmented slice-by-slice on the T1w image, avoiding subcutaneous fat and fascia (3D-slicer 4.4.0, https://www.slicer.org). Three approaches were applied to assess the diffusion metrics for each muscle:
i) Standard Tractography (STT): Manually, ROIs in the form of selection gates were drawn to segment the tracts of the upper leg muscles. This yielded sets of fiber tracts for each muscle, from which the mean FA, MD, RD and λ1 could be calculated using tract based analysis.
ii) Volume based Tractography (VBT): The manual segmentations were registered to the diffusion space using sequential rigid and b-spline transformations 7,8. The preprocessed diffusion data were split according to the segmentations and tractography was performed within the resulting segments of diffusion data. Diffusion metrics were assessed from the whole tracts for each muscle
iii) Manual segmentation based (MSB): The manual segmentations were smoothed and eroded by one voxel to avoid partial volume effects of non-muscular tissue and registered to the diffusion space to extract the diffusion metric for each muscle.
ANOVA analysis were performed using SPSS statistics to evaluate the main effect of muscles for the parameters FA, MD, RD and λ1.
Results
Figure 1 shows the different muscles for each method. Using STT large tract volumes were found, whereas VBT selects only the belly part of the muscle. MSB covers the whole muscle. The mean values of the diffusion parameters for each muscle and each methods are shown in Figure 2. For both tractbased analysis (STT and VBT) a significant (p<0.001) main effect of muscle was found for FA, MD, RD and λ1. No significant main effect was found when performing the manual segmentation without tractography (FA: p = 1; MD: p = 0.977; λ1 = 0.997; RD: p = 0.983).Discussion
Using fiber tractography it is
possible to visualize muscle fiber microstructure based on their diffusion properties.
Furthermore, tractography allows a weighting of tissue, since voxels which are
visited multiple times by calculated fibers, contribute more to the mean than
voxels which are affected by partial volume effects. Moreover, when performing
tractography, the FA range is limited between (0.1 – 0.6), since higher FA
values are most likely based on artifacts, but could also be an indication of
pathology, which would be excluded in this approach. In contrast to high variations
in MSB, the tractography methods show very similar results, even different
muscle volumes with different fiber count were observed. Although STT and VBT
are data driven approaches, they might have the prejudice of excluding
important muscle regions and VBT requires an accurate registration of a manual
annotation to diffusion space, which could become a source of errors.Conclusion
Our data suggest that using
tractography allows the assessment of intermuscular differences, which are
important in the context of disease progression monitoring and are missing when
evaluating data based on parameter maps.Acknowledgements
We thank Philips Germany and especially Burkhard Maedler for continuous scientific supportReferences
1. Schlaffke L, Rehmann R, Froeling M, et al. Diffusion tensor imaging of the human calf: Variation of inter- and intramuscle-specific diffusion parameters. J. Magn. Reson. Imaging 2017:1-12. doi:10.1002/jmri.25650.
2. Tax CMW, Otte WM, Viergever M a, Dijkhuizen RM, Leemans A. REKINDLE: Robust extraction of kurtosis INDices with linear estimation. Magn. Reson. Med. 2015;73(2):794-808. doi:10.1002/mrm.25165.
3. Veraart J, Sijbers J, Sunaert S, Leemans A, Jeurissen B. Weighted linear least squares estimation of diffusion MRI parameters: Strengths, limitations, and pitfalls. Neuroimage 2013;81:335-346. Available at: http://dx.doi.org/10.1016/j.neuroimage.2013.05.028.
4. Leemans A, Jones DK. The B-matrix must be rotated when correcting for subject motion in DTI data. Magn. Reson. Med. 2009;61(6):1336-49.
5. Irfanoglu MO, Walker L, Sarlls J, Marenco S, Pierpaoli C. Effects of image distortions originating from susceptibility variations and concomitant fields on diffusion MRI tractography results. Neuroimage 2012;61(1):275-288. doi:10.1016/j.neuroimage.2012.02.054.
6. Basser PJ, Pajevic S, Pierpaoli C, Duda J, Aldroubi A. In Vivo Fiber Tractography. Magnenetic Reson. Med. 2000;44:625-632.
7. Klein S, Staring M, Murphy K, Viergever M a., Pluim J. elastix: A Toolbox for Intensity-Based Medical Image Registration. IEEE Trans. Med. Imaging 2010;29(1):196-205. doi:10.1109/TMI.2009.2035616.
8. Shamonin D, Bron E, Lelieveldt B, Smits M, Klein S, Staring M. Fast parallel image registration on CPU and GPU for diagnostic classification of Alzheimer’s disease. Front. Neuroinform. 2014;7(January):1-15. doi:10.3389/fninf.2013.00050.
Figures
