Streamlets for visualisation and data exploration

TY - CONF

T1 - Streamlets for visualisation and data exploration

AU - Liptrot, Matthew George

PY - 2016/9/14

Y1 - 2016/9/14

N2 - Streamlets for visualisation and data exploration Matthew Liptrot, Image Group, Department of Computer Science, University of Copenhagen, Denmark Target Audience Anyone using streamlines for interpreting tractography from diffusion-weighted MRI Purpose The purpose of streamline tractography is to find connections between regions of the brain by integrating voxelwise information across a tensor field. However, simple quantification of propagation success, such as counting the number of streamlines between two end regions, are often combined with visualisation of all the streamlines to assess the spatial route taken is anatomically viable. Given the often huge number of streamlines generated in such experiments, and the 3D nature of the data, this visual inspection can be challenging. In addition, false interpretation of streamlines due to the natural perceptual association of a virtual streamline, which represents only a sample pathway through a tensor field, with a single in vivo brain fibre can be problematic. The recently proposed Streamlets method {Liptrot, 2015} addresses the latter issue, but here we also demonstrate how it can also benefit data visualisation and thereby also data exploration. Methods The Streamlets approach was applied to an example subject from the Human Connectome Project [Van Essen, 2013] using a standard installation of MRtrix v3 [Tournier, 2012]. Fibre directions were obtained using Constrained Spherical Deconvolution [Tournier, 2007], and 5 separate sets of streamlines were generated from every white-matter voxel using the iFOD2 algorithm [Tournier, 2010] included within MRtrix. For each set, the minimum and maximum streamline lengths were pre-specified, along with the FOD threshold below which tracking was halted. The parameters for the five sets were (min length mm : max length mm : FOD threshold): [16 : 300 : 0.9 / 8 : 100 : 0.7 / 4 : 9 : 0.5 / 2 : 5 : 0.3 / 1 : 3 : 0.1]. These arbitrary values were chosen to produce streamlines over a range of lengths, with longer ones requiring a contiguous chain of higher FOD thresholds to succeed. Visualisation of the streamlets was then done in MRtrix. Results The figures show the streamlets in a coronal slice through the mid-brain. The total number of streamlets was reduced by filtering, and by removing any that did not pass through the chosen slice. Results are colour coded either by direction, following the standard DEC-FA colour scheme (red: left-right, green: anterior-posterior, blue: inferior-superior), or with a separate colour per streamlet set. A non-diffusion-weighted image is used as background. The separation of known pathways according to streamlet set is very apparent (Figs 1 & 2). In Figure 3, one can see how the reduction in number and length of streamlines permits a visual appreciation of the undulating character of the pathways found across the corpus callosum which was recently reported elsewhere [Dyrby, 2014]. Discussion The clear separation of streamline sets according to tract membership, as seen in Figs 1 & 2, offers assurance that the tractography is indeed behaving as expected. In addition, it is more obvious how the more confident streamlets occupy the central parts of the tracts, which is also to be expected given the likelihood for partial volume effects towards a tract’s outer shell. The lack of streamlets projecting laterally from the corpus callosum also indicates the difficulty in crossing the complexities in the corona radiata. And the visualisation of undulating streamlets in the corpus callosum (Figure 3) demonstrates how useful the technique is for uncovering data features that are often easy to overlook. Conclusion Streamlets can be a useful tool for quality-assessment of streamline tractography, and permit a deeper look into the spatial pathways that they follow. In particular, some features become more apparent which could lead to better understanding of tractographic results. References Dyrby, T et al. Undulating and crossing axons in the corpus callosum may explain the overestimation of axon diameters with ActiveAx, Proc ISMRM, 2014 Van Essen, D et al. The WU-Minn Human Connectome Project: An overview. NeuroImage 80 (2013) :62-79. Liptrot, M.G., Streamlets: Preventing Over-Interpretation of Streamlines, Proceedings of Human Brain Mapping conference, 2015. Tournier, J.-D.; Calamante, F. & Connelly, A. Robust determination of the fibre orientation distribution in diffusion MRI: Non-negativity constrained super-resolved spherical deconvolution. NeuroImage, 2007, 35, 1459-1472 Tournier, J.-D. et al Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distributions. Proc ISMRM, 2010, 1670 Tournier, J.-D. et al. MRtrix: Diffusion tractography in crossing fiber regions. Int. J. Imaging Syst. Technol., 2012, 22, 53-66

AB - Streamlets for visualisation and data exploration Matthew Liptrot, Image Group, Department of Computer Science, University of Copenhagen, Denmark Target Audience Anyone using streamlines for interpreting tractography from diffusion-weighted MRI Purpose The purpose of streamline tractography is to find connections between regions of the brain by integrating voxelwise information across a tensor field. However, simple quantification of propagation success, such as counting the number of streamlines between two end regions, are often combined with visualisation of all the streamlines to assess the spatial route taken is anatomically viable. Given the often huge number of streamlines generated in such experiments, and the 3D nature of the data, this visual inspection can be challenging. In addition, false interpretation of streamlines due to the natural perceptual association of a virtual streamline, which represents only a sample pathway through a tensor field, with a single in vivo brain fibre can be problematic. The recently proposed Streamlets method {Liptrot, 2015} addresses the latter issue, but here we also demonstrate how it can also benefit data visualisation and thereby also data exploration. Methods The Streamlets approach was applied to an example subject from the Human Connectome Project [Van Essen, 2013] using a standard installation of MRtrix v3 [Tournier, 2012]. Fibre directions were obtained using Constrained Spherical Deconvolution [Tournier, 2007], and 5 separate sets of streamlines were generated from every white-matter voxel using the iFOD2 algorithm [Tournier, 2010] included within MRtrix. For each set, the minimum and maximum streamline lengths were pre-specified, along with the FOD threshold below which tracking was halted. The parameters for the five sets were (min length mm : max length mm : FOD threshold): [16 : 300 : 0.9 / 8 : 100 : 0.7 / 4 : 9 : 0.5 / 2 : 5 : 0.3 / 1 : 3 : 0.1]. These arbitrary values were chosen to produce streamlines over a range of lengths, with longer ones requiring a contiguous chain of higher FOD thresholds to succeed. Visualisation of the streamlets was then done in MRtrix. Results The figures show the streamlets in a coronal slice through the mid-brain. The total number of streamlets was reduced by filtering, and by removing any that did not pass through the chosen slice. Results are colour coded either by direction, following the standard DEC-FA colour scheme (red: left-right, green: anterior-posterior, blue: inferior-superior), or with a separate colour per streamlet set. A non-diffusion-weighted image is used as background. The separation of known pathways according to streamlet set is very apparent (Figs 1 & 2). In Figure 3, one can see how the reduction in number and length of streamlines permits a visual appreciation of the undulating character of the pathways found across the corpus callosum which was recently reported elsewhere [Dyrby, 2014]. Discussion The clear separation of streamline sets according to tract membership, as seen in Figs 1 & 2, offers assurance that the tractography is indeed behaving as expected. In addition, it is more obvious how the more confident streamlets occupy the central parts of the tracts, which is also to be expected given the likelihood for partial volume effects towards a tract’s outer shell. The lack of streamlets projecting laterally from the corpus callosum also indicates the difficulty in crossing the complexities in the corona radiata. And the visualisation of undulating streamlets in the corpus callosum (Figure 3) demonstrates how useful the technique is for uncovering data features that are often easy to overlook. Conclusion Streamlets can be a useful tool for quality-assessment of streamline tractography, and permit a deeper look into the spatial pathways that they follow. In particular, some features become more apparent which could lead to better understanding of tractographic results. References Dyrby, T et al. Undulating and crossing axons in the corpus callosum may explain the overestimation of axon diameters with ActiveAx, Proc ISMRM, 2014 Van Essen, D et al. The WU-Minn Human Connectome Project: An overview. NeuroImage 80 (2013) :62-79. Liptrot, M.G., Streamlets: Preventing Over-Interpretation of Streamlines, Proceedings of Human Brain Mapping conference, 2015. Tournier, J.-D.; Calamante, F. & Connelly, A. Robust determination of the fibre orientation distribution in diffusion MRI: Non-negativity constrained super-resolved spherical deconvolution. NeuroImage, 2007, 35, 1459-1472 Tournier, J.-D. et al Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distributions. Proc ISMRM, 2010, 1670 Tournier, J.-D. et al. MRtrix: Diffusion tractography in crossing fiber regions. Int. J. Imaging Syst. Technol., 2012, 22, 53-66

KW - Faculty of Science

M3 - Poster

T2 - ISMRM 2016 Workshop on Breaking the Barriers of Diffusion MRI

Y2 - 11 September 2016 through 16 September 2016

ER -