Enhancing Tractometry Robustness with Streamline and Bundle-Specific Nonlinear Registration
Poster No:
1601
Submission Type:
Abstract Submission
Authors:
Bramsh Chandio1, Sophia Thomopoulos2, Jaroslaw Harezlak3, Eleftherios Garyfallidis4, Paul Thompson5
Institutions:
1University of Southern California, Los Angeles, CA, 2USC, Marina del Rey, CA, 3Indiana University School of Public Health, Bloomington, IN, 4Indiana University Bloomington, Bloomington, IN, 5USC, Marina Del Rey, CA
First Author:
Co-Author(s):
Introduction:
In tractometry, the goal is to analyze microstructural measures projected on specific segments along white matter tracts. Registration is used to establish correspondences between these segments across subjects, enabling segment-specific analyses. We incorporate BundleWarp[1] nonlinear streamline-specific bundle registration into BUndle ANalytics (BUAN)[2] tractometry pipeline to ensure that white matter tracts are accurately aligned and can be effectively analyzed and compared across individuals, enhancing the reliability and interpretability of results. We evaluate the effects of BundleWarp registration on group statistics using Alzheimer's Disease Neuroimaging Initiative Phase 3 (ADNI3) data[3].
Methods:
We add BundleWarp registration into the BUAN tractometry pipeline, as shown in Figure 1. BUAN extracts bundles from populations and analyzes the microstructural measures (e.g., FA) projected onto the bundles along the length of the tracts to test for significant group differences. BUAN creates N horizontal segments along the length of the tracts to be analyzed. Horizontal segments are created based on points on the streamlines of a bundle belonging to the closest model bundle centroid point. Here, we add a partially deformable BundleWarp registration to nonlinearly align subjects' bundles with model bundles before creating the horizontal segments to find better segment correspondences among populations. Linear Mixed Models are applied to 6 WM bundles with group, age, and sex modeled as fixed effects and subject as a random effect term, the response variable being each DTI metric. We corrected for multiple comparisons using the FDR method. Note that the nonlinearly moved bundles are only used to assign segment numbers to points on the streamlines in the bundles. The actual statistical analysis always takes place in the native space of diffusion data. This step uses bundles of the original shape and microstructural measures in the native space, using segment labels given during the assignment step for segment-specific group analysis.
Results:
Fig. 2 shows BUAN results (with and without BundleWarp) on the AF_L bundle using the FA metric. Fig.1 bottom panel, overall, from the results on six bundles (AF_L, AF_R, CST_L, CST_R, IFOF_L, and IFOF_R), BundleWarp made BUAN robust to outliers due to misaligned segments across groups and subjects . It enhanced the sensitivity for detecting group differences by mitigating errors from anatomical misregistration across subjects. However, in specific instances, a decrease in significance indicates that certain confounding effects may have been corrected that were due to disparities in tract alignment within each group. Moreover, adding nonlinear registration did not cause artifacts that substantially alter results relative to those obtained without BundleWarp.
Conclusions:
BundleWarp is a valuable addition to the BUAN tractometry pipeline for robust tractometric analysis. Nonlinearly aligning subjects' WM bundles with model bundles improves along-tract segment correspondence, increasing the sensitivity of group statistical analyses by eliminating errors due to misalignment across subjects.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 2
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis 1
Image Registration and Computational Anatomy
Methods Development
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity
Keywords:
Aging
Data Registration
Degenerative Disease
Design and Analysis
Machine Learning
Open-Source Software
Tractography
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
Other
1|2Indicates the priority used for review
Provide references using author date format
[2] Chandio, B.Q., Risacher, S.L., Pestilli, F., Bullock, D., Yeh, F.C., Koudoro, S., Rokem, A., Harezlak, J. and Garyfallidis, E., 2020. Bundle analytics, a computational framework for investigating the shapes and profiles of brain pathways across populations. Scientific Reports, 10(1), pp.1-18.
[3] Zavaliangos-Petropulu, A., Nir, T.M., Thomopoulos, S.I., Reid, R.I., Bernstein, M.A., Borowski, B., Jack Jr, C.R., Weiner, M.W., Jahanshad, N. and Thompson, P.M., 2019. Diffusion MRI indices and their relation to cognitive impairment in brain aging: the updated multi-protocol approach in ADNI3. Frontiers in Neuroinformatics, 13, p.2.
[4] Thomopoulos, S.I., Nir, T.M., Villalon-Reina, J.E., Zavaliangos-Petropulu, A., Maiti, P., Zheng, H., Nourollahimoghadam, E., Jahanshad, N., Thompson, P.M., for the Alzheimer’s Disease Neuroimaging Initiative. Diffusion MRI Metrics and their Relation to Alzheimer’s Disease Severity: Effects of Harmonization Approaches. SIPAIM 2021, Campinas, Brazil. 2021.
[5] 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 35, 1459–1472 (2007).
[6] Girard, G., Whittingstall, K., Deriche, R., & Descoteaux, M. Towards quantitative connectivity analysis: reducing tractography biases. NeuroImage, 98, 266-278, 2014.
[7] Garyfallidis, E., Brett, M., Amirbekian, B., Rokem, A., van der Walt, S., Descoteaux, M. and Nimmo-Smith, I., “Dipy, a library for the analysis of diffusion MRI data,” Front. Neuroinform. 8, 8 (2014).
[8] Garyfallidis, E., Côté, M.A., Rheault, F., Sidhu, J., Hau, J., Petit, L., Fortin, D., Cunanne, S. and Descoteaux, M., 2018. Recognition of white matter bundles using local and global streamline-based registration and clustering. NeuroImage, 170, pp.283-295.