Print Close

Along-Tract Analysis of White Matter Bundles: Comparing Volumetric and Streamline-based Approaches

Poster No:

1849

Submission Type:

Abstract Submission

Authors:

Iyad Ba Gari¹, Siddharth Narula¹, Shruti Gadewar¹, Bramsh Chandio¹, Paul Thompson¹, Talia Nir¹, Neda Jahanshad¹

Institutions:

¹Imaging Genetics Center, Keck School of Medicine of University of Southern California, Los Angeles, California

First Author:

Iyad Ba Gari
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Co-Author(s):

Siddharth Narula
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Shruti Gadewar
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Bramsh Chandio
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Paul Thompson, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Talia Nir, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Neda Jahanshad, PhD
Imaging Genetics Center, Keck School of Medicine of University of Southern California
Los Angeles, California

Introduction:

Along-tract analysis of white matter (WM) bundles can help to map detailed patterns of WM pathway degeneration in Alzheimer's disease and related dementias [1]. Common along-tract methods are often based on simplifying subject bundles to one centroid streamlines and more advanced methods such as Bundle Analytics (BUAN) analyze all possible streamlines in a particular bundle to take anatomical variations into account [2]. While ideal for identifying subtle variations, streamline methods can be computationally intensive, with computation time scaling by the number of streamlines and sample size. As reliable neuroimaging studies require large sample sizes, faster and scalable methods are needed. Here, we present our novel Medial Tractography Analysis (MeTA) method. MeTA aims to minimize partial voluming and microstructural heterogeneity in diffusion MRI (dMRI) metrics by extracting and parcellating the core volume along the bundle length in the voxel-space directly while effectively preserving bundle shape and efficiently capturing the regional variation within and along WM bundles [3]. We evaluated along-tract WM microstructure associations with cognitive impairment using MeTA compared to the established streamline-based BUAN method. We also compared performance efficiency across both methods.

Methods:

We used dMRI and clinical data from 714 participants (aged 55-96 years, 52% females) enrolled in ADNI3 [4]. The dMRI data were processed using the following steps: denoised using Local PCA [5], deGibbs, and corrected for eddy current and motion artifacts using FSL eddy [6]. The fiber orientation distributions (iFOD2) were estimated using a multi-tissue constrained spherical deconvolution approach [7]. Whole brain probabilistic streamline tractography was generated using the probabilistic iFOD2 method [8]. We used RecoBundles [9] to identify and segment 12 WM bundles. We applied two tractometry methods, MeTA and BUAN, to evaluate the diffusion tensor (DTI) fractional anisotropy (FA) and mean diffusivity (MD) associations with 1) clinical impairment (N=274) compared to cognitively normal (CN) individuals (N=440); and 2) Clinical Dementia Rating Scale Sum of Boxes (CDR-SB). Full details of the tractometry pipelines can be found in Figure 1. We used linear mixed models, where age and sex were modeled as fixed effects, and subjects (as required for BUAN's streamline-based analyses) and scan sites were modeled as nested random variables. We corrected for multiple comparisons across 315 tests (21 bundles x 15 regions) using the false discovery rate (FDR) procedure (q=0.05). We note that BUAN has traditionally been run across 100 segments of a bundle, but we analyze 15 for consistency between the two methods.

Results:

MeTA completed statistical tests across 15 segments in a single bundle in about 15 seconds, using 20 MB of memory. In contrast, BUAN took around 15 hours and required 70 GB of memory (expected as it uses rich definitions of bundles). Across both methods participants with cognitive impairment and higher CDR-SB scores showed negative associations with FA values and positive associations with MD measures across bundles such as AF_L, CPH_R, IFOF_R, CST, UF_R, and CC as displayed in Figure 2. However, DTI metrics along the length of the bundles had stronger regional associations using MeTA compared to BUAN.

Conclusions:

MeTA is a fast and reliable approach for identifying regional WM brain abnormalities in clinical populations at both the subject and group levels. ADRD associations detected with MeTa were largely consistent with existing along-tract pipelines. While it does not analyze individual streamlines and requires a coarser along-tract parcellation scheme, MeTA is much faster than current state-of-the-art streamline methods with compatible, if not more sensitive findings. Such highly scalable approaches can be complementary to existing methods and applied to large-scale efforts and consortia.

Disorders of the Nervous System:

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ²

Modeling and Analysis Methods:

Diffusion MRI Modeling and Analysis

Methods Development ¹

Keywords:

Statistical Methods

Tractography

White Matter

Other - Medial Tractography Analysis

^1|2Indicates the priority used for review

Provide references using author date format

[1] A. Pichet Binette et al., “Bundle-specific associations between white matter microstructure and Aβ and tau pathology in preclinical Alzheimer’s disease,” Elife, vol. 10, May 2021, doi: 10.7554/eLife.62929.
[2] B. Q. Chandio et al., “Bundle analytics, a computational framework for investigating the shapes and profiles of brain pathways across populations,” Sci. Rep., vol. 10, no. 1, p. 17149, Oct. 2020.
[3] I. Ba Gari et al., “Along-Tract Parameterization of White Matter Microstructure using Medial Tractography Analysis (MeTA),” in The 19th International Symposium on Medical Information Processing and Analysis, 2023.
[4] A. Zavaliangos-Petropulu et al., “Diffusion MRI Indices and Their Relation to Cognitive Impairment in Brain Aging: The Updated Multi-protocol Approach in ADNI3,” Front. Neuroinform., vol. 13, p. 2, Feb. 2019.
[5] J. V. Manjón, P. Coupé, L. Concha, A. Buades, D. L. Collins, and M. Robles, “Diffusion weighted image denoising using overcomplete local PCA,” PLoS One, vol. 8, no. 9, p. e73021, Sep. 2013.
[6] J. L. R. Andersson and S. N. Sotiropoulos, “An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging,” Neuroimage, vol. 125, pp. 1063–1078, Jan. 2016.
[7] B. Jeurissen, J.-D. Tournier, T. Dhollander, A. Connelly, and J. Sijbers, “Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data,” Neuroimage, vol. 103, pp. 411–426, Dec. 2014.
[8] J. D. Tournier, F. Calamante, A. Connelly, and Others, “Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distributions,” in Proceedings of the international society for magnetic resonance in medicine, John Wiley & Sons, Inc, New Jersey, 2010. Accessed: Dec. 15, 2022. [Online]. Available: https://archive.ismrm.org/2021/1767.html
[9] E. Garyfallidis et al., “Recognition of white matter bundles using local and global streamline-based registration and clustering,” Neuroimage, vol. 170, pp. 283–295, Apr. 2018.
[10] P. A. Yushkevich, H. Zhang, and J. C. Gee, “Continuous medial representation for anatomical structures,” IEEE Trans. Med. Imaging, vol. 25, no. 12, pp. 1547–1564, Dec. 2006.

Acknowledgments:
This research was supported by NIH grants P41EB015922, R01AG059874, R01MH134004, and RF1AG057892. Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and im