Concurrent Mapping of Brain Connections in Neonates and Adults with dMRI Tractography
Poster No:
2368
Submission Type:
Abstract Submission
Authors:
Wei Zhang1, Yijie Li1, Ruixi Zheng1, Yuqian Chen2, Leo Zekelman2, Jarrett Rushmore3, Yogesh Rathi2, Nikos Makris2, Lauren O'Donnell2, Fan Zhang1
Institutions:
1University of Electronic Science and Technology of China, Chengdu, Sichuan, 2Harvard Medical School, Boston, MA, 3Boston University, Boston, MA
First Author:
Co-Author(s):
Introduction:
Comparison of the white matter (WM) structural connections between adult and neonate brains via diffusion MRI (dMRI) can contribute to our understanding of normal brain development and assist identification of potential biomarkers for neurological disorders1,2. Brain WM atlases are an important tool to reveal WM differences between adults and neonates3,4,5. However, there have been no atlases that can concurrently map WM connections across adult and neonate populations.
This study presents a novel dMRI tractography streamline clustering atlas consisting of WM connections from both adults and newborns. The atlas is created concurrently using the high-quality dMRI data from the Human Connectome Project (HCP)6 and the Developing Human Connectome Project (dHCP)7, where each cluster in the atlas includes streamlines from both populations. With the proposed atlas, we perform a comparative study of major WM fiber bundles (n=58) between the two populations and find widespread differences in WM structures across the brain.
This study presents a novel dMRI tractography streamline clustering atlas consisting of WM connections from both adults and newborns. The atlas is created concurrently using the high-quality dMRI data from the Human Connectome Project (HCP)6 and the Developing Human Connectome Project (dHCP)7, where each cluster in the atlas includes streamlines from both populations. With the proposed atlas, we perform a comparative study of major WM fiber bundles (n=58) between the two populations and find widespread differences in WM structures across the brain.
Methods:
Dataset: dMRI data from 100 healthy young adults (29±7 years; 50 F, 50 M) from the HCP database and 100 healthy new-born babies (33.43±8.86 post-conceptional weeks; 50 F, 50 M) were used.
Atlas creation: Fig 1 gives the method overview. First, whole-brain tractography is performed to reconstruct whole brain WM streamlines. We use our two-tensor Unscented Kalman filter (UKF) algorithm8, which accounts for crossing fibers and offers sensitive and reliable fiber tracking within a wide range of populations5,9. Second, streamline clustering is performed to subdivide the tractography into fine-scale clusters. We use our well-established whitematteranalysis pipeline5 for simultaneous streamline clustering across all HCP and dHCP subjects. This generates a streamline clustering atlas including 800 clusters. Third, anatomical atlas curation is performed to annotate each cluster with an anatomical label according to our ORG atlas5,10. We calculate cluster distances between the new atlas and the ORG atlas, and then assign each new atlas cluster with the label of the closest ORG cluster. In total, our new atlas contains an anatomical tract parcellation including 58 major anatomical WM tracts. Finally, subject-specific tractography parcellation is performed by applying the curated atlas to the whole brain tractography of each individual in the HCP and dHCP datasets.
Comparison between HCP and dHCP: Our atlas was created from both HCP and dHCP, with each cluster containing streamlines from both populations. This ensures that the atlas clustering is consistent across the two populations. We perform two types of comparisons. First, we measure the coefficient of variation (CoV) of the number of streamlines (NoF) of each tract in each population. This is to compare the variability of the corresponding WM tracts between the two populations. Second, we measure the lateralization index (LI) of NoF of each association tract in each population. This is to compare the differences of WM lateralization effect between the two populations.
Atlas creation: Fig 1 gives the method overview. First, whole-brain tractography is performed to reconstruct whole brain WM streamlines. We use our two-tensor Unscented Kalman filter (UKF) algorithm8, which accounts for crossing fibers and offers sensitive and reliable fiber tracking within a wide range of populations5,9. Second, streamline clustering is performed to subdivide the tractography into fine-scale clusters. We use our well-established whitematteranalysis pipeline5 for simultaneous streamline clustering across all HCP and dHCP subjects. This generates a streamline clustering atlas including 800 clusters. Third, anatomical atlas curation is performed to annotate each cluster with an anatomical label according to our ORG atlas5,10. We calculate cluster distances between the new atlas and the ORG atlas, and then assign each new atlas cluster with the label of the closest ORG cluster. In total, our new atlas contains an anatomical tract parcellation including 58 major anatomical WM tracts. Finally, subject-specific tractography parcellation is performed by applying the curated atlas to the whole brain tractography of each individual in the HCP and dHCP datasets.
Comparison between HCP and dHCP: Our atlas was created from both HCP and dHCP, with each cluster containing streamlines from both populations. This ensures that the atlas clustering is consistent across the two populations. We perform two types of comparisons. First, we measure the coefficient of variation (CoV) of the number of streamlines (NoF) of each tract in each population. This is to compare the variability of the corresponding WM tracts between the two populations. Second, we measure the lateralization index (LI) of NoF of each association tract in each population. This is to compare the differences of WM lateralization effect between the two populations.
Results:
Fig 2a gives a visualization of example fiber tracts, showing while many tracts have visually different shapes between HCP and dHCP (e.g. AF and MCP), some are highly visually similar (e.g. CST and ILF). Fig 2b shows that dHCP has generally higher CoV of NoF across the fiber tracts, indicating a higher variability of the brain WM structures in the newborns than the adults. Fig 3c shows the lateralization effects in newborns are generally smaller than in adults. Interestingly, the AF tract has the highest LI scores across all association tracts in the atlas,where HCP has a stronger lateralization effect than dHCP.
Conclusions:
This study presents a novel cross-population WM atlas for concurrent mapping of brain connections between adults and neonates. We perform a comparative assessment of the WM connections and identify widespread differences between the two populations in terms of streamline counts and tract visualization.
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity 2
Neuroinformatics and Data Sharing:
Brain Atlases
Novel Imaging Acquisition Methods:
Diffusion MRI 1
Keywords:
Atlasing
Segmentation
Tractography
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review
Provide references using author date format
2. Warrington, S. (2022), 'Concurrent mapping of brain ontogeny and phylogeny within a common space: Standardized tractography and applications', Science Advances, 8(42), eabq2022.
3. Wilkinson, M. (2017), 'Detection and Growth Pattern of Arcuate Fasciculus from Newborn to Adult', Frontiers in Neuroscience, 11, 389.
4. Liang, W. (2022), 'A comparative study of the superior longitudinal fasciculus subdivisions between neonates and young adults', Brain Structure & Function, 227(8), 2713–2730.
5. Zhang, F. (2018), 'An anatomically curated fiber clustering white matter atlas for consistent white matter tract parcellation across the lifespan', NeuroImage, 179, 429–447.
6. Elam, J. S. (2021), 'The Human Connectome Project: A retrospective', NeuroImage, 244, 118543.
7. Makropoulos, A. (2017), 'The Developing Human Connectome Project: a Minimal Processing Pipeline for Neonatal Cortical Surface Reconstruction', In bioRxiv (p. 125526). https://doi.org/10.1101/125526
8. Malcolm, J. G. (2010), 'Filtered multitensor tractography', IEEE Transactions on Medical Imaging, 29(9), 1664–1675.
9. Zhang, F. (2020), 'Deep white matter analysis (DeepWMA): Fast and consistent tractography segmentation', Medical Image Analysis, 65, 101761.
10. Zhang, F. (2019). 'Test-retest reproducibility of white matter parcellation using diffusion MRI tractography fiber clustering', Human Brain Mapping, 40(10), 3041–3057.