Enhanced Structural Brain Connectivity Analyses Using High Diffusion-weighting Strengths
Poster No:
2361
Submission Type:
Abstract Submission
Authors:
Leyao Yu1, Adeen Flinker2,1, Jelle Veraart2
Institutions:
1New York University Tandon School of Engineering, New York, NY, 2New York University Grossman School of Medicine, New York, NY
First Author:
Co-Author(s):
Adeen Flinker
New York University Grossman School of Medicine|New York University Tandon School of Engineering
New York, NY|New York, NY
New York University Grossman School of Medicine|New York University Tandon School of Engineering
New York, NY|New York, NY
Introduction:
Tractography using diffusion-weighted Magnetic Resonance Imaging (dMRI) is an emerging tool in clinical and neuroscientific research, enabling presurgical planning (Nimsky et al., 2007) and the study of structural connectivity networks (Behrens & Sporns, 2012, Duffau, 2008; Glasser & Rilling, 2008; Turken & Dronkers, 2011). Recent studies showed that with increasing diffusion-weighting strength, probed by the b-value, dMRI becomes primarily sensitive to the intra-axonal signal (Veraart et al., 2016, 2020). Thus, increasing b-values may hold the untapped potential to increase the reliability of tractography. Here we test the hypothesis that exceeding current standard clinical b-values improve reliable reconstruction of longer-range connections using dMRI-based tractography, quantified based on nominal values and graph network metrics (Rubinov & Sporns, 2010).
Methods:
Data: Multishell dMRI data was used from the publicly available MICRA dataset (Koller et al., 2021) where diffusion weighting was acquired with b=200, 500, 1200, 2400, and 6000 s/mm2 from 6 healthy participants with 5 repetitions.
Analysis: Whole-brain, probabilistic tractography is performed using anatomical constraints (Tournier et al., 2019). The MRtrix3.0 package is used for fiber orientation estimation and tractography. Connectivity matrices (68 x 68 Desikan-Killiany (Desikan et al., 2006) nodes were constructed by computing the count and density of streamlines that connect corresponding cortical areas. Commonly used graph-theoretical metrics (average network degree and strength, betweenness, centrality, clustering coefficient, and normalized global efficiency) were calculated with the igraph toolbox in R.
Analysis: Whole-brain, probabilistic tractography is performed using anatomical constraints (Tournier et al., 2019). The MRtrix3.0 package is used for fiber orientation estimation and tractography. Connectivity matrices (68 x 68 Desikan-Killiany (Desikan et al., 2006) nodes were constructed by computing the count and density of streamlines that connect corresponding cortical areas. Commonly used graph-theoretical metrics (average network degree and strength, betweenness, centrality, clustering coefficient, and normalized global efficiency) were calculated with the igraph toolbox in R.
Results:
(i) More robust tracking of long-range connections (Fig 1): We first characterized the global and local change in the number and length of streamlines in long/short term range connections by fitting a polynomial mixed effect model. We found the total number of interhemispheric connections increased from 4.4% to 7.0% from b= 1200 to 6000s/mm2. Focusing on critical regions for the language network, we found that the average length of streamlines connecting the Inferior Frontal Gyrus (IFG) and Middle Frontal Gyrus (MFG) significantly increased by 24.08% and 34.70% in the same b-range. This is further reflected via increment in meaningful long-range nodal-to-nodal connections. For example, the number of streamlines seeded from left IFG to left STG increased by 160.6%. Further, spurious short-range connections and self-connections decreased with increasing b-value. (mean decrease in six major language ROIs -23.04%, std 7.76%; pars opercularis to caudal MFG -69.87%, MTG to STG -25.83%, all in left hemisphere).
(ii) Impact on network analysis (Fig 2): We then characterized the changes of global and nodal graph network metrics by fitting a polynomial mixed effect model. Global changes in centrality of streamline count is limited to -13.79%, while certain regions are more vulnerable to b-value changes. In left STG and left MTG, the centrality decreases dramatically, (STG -48.27% and MTG -29.49%, from b= 1200 to 6000s/mm2), reflecting the heterogeneity of white matter tracts in the cortex.
(iii) No loss in reproducibility: To quantify reproducibility, we calculated the coefficient of variation (CoV) and intraclass correlation coefficient for global and local networks. Using ANOVAN analysis, we do not observe loss in reproducibility of the findings for b-values above 1200s/mm2, despite the lower SNR at high b.
(ii) Impact on network analysis (Fig 2): We then characterized the changes of global and nodal graph network metrics by fitting a polynomial mixed effect model. Global changes in centrality of streamline count is limited to -13.79%, while certain regions are more vulnerable to b-value changes. In left STG and left MTG, the centrality decreases dramatically, (STG -48.27% and MTG -29.49%, from b= 1200 to 6000s/mm2), reflecting the heterogeneity of white matter tracts in the cortex.
(iii) No loss in reproducibility: To quantify reproducibility, we calculated the coefficient of variation (CoV) and intraclass correlation coefficient for global and local networks. Using ANOVAN analysis, we do not observe loss in reproducibility of the findings for b-values above 1200s/mm2, despite the lower SNR at high b.
Conclusions:
We observe that b-value is a critical experimental design factor that impacts tractography and structural connectivity analyses. Higher b-values are more robust in reconstructing long range connections, less impacted by spurious short-range connections, thereby advancing the study of structural brain connectivity.
Language:
Language Other 2
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis
Novel Imaging Acquisition Methods:
Diffusion MRI 1
Keywords:
Language
Modeling
Tractography
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review
Provide references using author date format
Desikan, R. S. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage, 31(3), 968–980.
Duffau, H. (2008). The anatomo-functional connectivity of language revisited. New insights provided by electrostimulation and tractography. Neuropsychologia, 46(4), 927–934.
Glasser, M. F. (2008). DTI tractography of the human brain’s language pathways. Cerebral Cortex , 18(11), 2471–2482.
Koller, K. (2021). MICRA: Microstructural image compilation with repeated acquisitions. NeuroImage, 225, 117406.
Nimsky, C. (2007). Preoperative and intraoperative diffusion tensor imaging-based fiber tracking in glioma surgery. Neurosurgery, 61(1 Suppl), 178–185; discussion 186.
Rubinov, M. (2010). Complex network measures of brain connectivity: uses and interpretations. NeuroImage, 52(3), 1059–1069.
Tournier, J. D. (2019). MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. NeuroImage, 202, 116137.
Turken, A. U. (2011). The neural architecture of the language comprehension network: converging evidence from lesion and connectivity analyses. Frontiers in Systems Neuroscience, 5, 1.
Veraart, J. (2016). Universal power-law scaling of water diffusion in human brain defines what we see with MRI.
Veraart, J. (2020). Noninvasive quantification of axon radii using diffusion MRI. eLife, 9. https://doi.org/10.7554/eLife.49855