Manon Edde¹, Etienne St-Onge¹, Antoine Théberge^1,2, Guillaume Theaud¹, Emmanuelle Renauld¹, and Maxime Descoteaux^1
1Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, QC, Canada, ²Videos & Images Theory and Analytics Laboratory (VITAL), Université de Sherbrooke, Sherbrooke, QC, Canada

Tractography algorithms require a seeding map in which each point initiates one streamline. We can determine how many streamlines end in each voxel from one of their ends, but their initial point is unknown. We propose a method to compute the initial streamline points (IPS) in each voxel and evaluate their distribution depending on two strategies of seeding. With the surface-based interface, IPSs maps are more uniform along the cortex independently of the cortical volume, allowing a greater cortical coverage and specificity at the level of the cortical regions both at the whole-brain and bundle level.

Alessandro Daducci¹, Francesco Gobbi¹, Nicola Febbrari¹, Matteo Battocchio¹, and Simona Schiavi^1
1University of Verona, Verona, Italy

Tractography is a powerful tool to study brain connectivity but it suffers from an intrinsic trade-off between sensitivity and specificity. The former can be increased by constructing more streamlines, while filtering techniques can improve the latter. However, creating many streamlines may introduce redundancy in the tractograms and negatively affect the performances of filtering methods, especially those based on linear optimization. Here, we present the “blurred streamlines”, a novel concept based on a combination of streamline clustering and spatial blurring of their signal contributions. Preliminary results show the potential of this formulation and open new perspectives for improving tractography accuracy.

Viljami Sairanen^1,2 and Chantal M. W. Tax^3,4
1Department of Computer Science, University of Verona, Verona, Italy, ²Translational Imaging in Neurology, Department of Medicine and Biomedical Engineering, University of Basel, Basel, Switzerland, ³CUBRIC, Cardiff University, Cardiff, United Kingdom, ⁴Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

Diffusion MRI measurements are affected by noise which propagates into the model estimation. Residual bootstrapping has been proposed to assess the uncertainty of parameter estimates, but do not consider the confounding effect of measurement outliers (e.g. due to subject motion), limiting their usage on clinical data. We present a robust bootstrapping algorithm and demonstrate its performance on the estimation and uncertainty quantification of rotationally invariant spherical harmonic (RISH) features with simulations and clinical data. Our algorithm can improve the reliability of cross-scanner harmonization relying on RISH and probabilistic tractography.

Patryk Filipiak¹, Lee Basler², Anthony Zuccolotto², Ying-Chia Lin¹, Dimitris G. Placantonakis³, Timothy Shepherd¹, Walter Schneider⁴, Fernando E. Boada¹, and Steven H. Baete^1
1Center for Advanced Imaging Innovation and Research (CAI2R), NYU School of Medicine, New York, NY, United States, ²Psychology Software Tools, Inc., Pittsburgh, PA, United States, ³Department of Neurosurgery, Perlmutter Cancer Center, Neuroscience Institute, Kimmel Center for Stem Cell Biology, NYU School of Medicine, New York, NY, United States, ⁴University of Pittsburgh, Pittsburgh, PA, United States

ODF peak finding typically fails to reconstruct fibers crossing at angles below 45 degrees. We aim to break this barrier with ODF-Fingerprinting. Our approach replaces the peak finding mechanism with pattern matching, allowing to use all the information stored in ODFs. In this work, we study the ability of ODF-Fingerprinting to reconstruct fibers crossing at 90, 45, and 30 degrees in a diffusion phantom composed of textile tubes with 0.8µm diameter, approaching the anatomical scale of axons. Our approach reaches much higher sensitivity (84-100%) than the ODF peak finding (0-33%), especially at the shallow angles. 

Hunter Moss¹ and Jens Jensen^1
1Neurosciences, Medical University of South Carolina, Charleston, SC, United States

The fiber orientation distribution function (fODF) can be estimated with diffusion MRI and gives the angular probability density of axon orientations in white matter. Here fODFs, as estimated with constrained spherical deconvolution (CSD) and fiber ball imaging (FBI), are compared quantitatively in order to assess their consistency. An important distinction between these two approaches is that CSD relies on a globally defined, empirical response function while FBI does not. For two measures of anisotropy, similar results are found for the CSD and FBI fODFs. However, a distance metric reveals significant differences in fODF peak directions and fine structure.

Zifei Liang¹ and Jiangyang Zhang^1
1Center for Biomedical Imaging, Dept. of Radiology, New York University School of Medicine, NEW YORK, NY, United States

Diffusion MRI based tractography is widely used to examine structural connectivity in the brain. Due to noises and motions, tractography in individual subject may contain erroneous results, which may be removed by averaging over a group. Although several group average tractography (GAT) approaches are available, their accuracy has not been thoroughly examined. In this study, we compared GAT based on spatial normalization of fiber orientation distribution maps and direct streamline mapping. Our results suggest that direct streamline mapping better preserve small and secondary axonal projections and is better-suited for studying group average tractography of the brain.

Bramsh Qamar Chandio¹ and Eleftherios Garyfallidis^1
1Intelligent Systems Engineering, Indiana University Bloomington, Bloomington, IN, United States

StND is a method for non-rigid registration of white matter tracts in streamline-space. StND performs two-step registration to align the moving bundle with a static (reference) bundle. In the first step, it affinely registers the two given tracts, and in the second step, it partially deforms the moving bundle to better align it with the static bundle. This partial deformation allows us to improve registration by deforming only areas where there is higher correspondence between streamlines of the tracts. This also helps in preserving the original anatomical structure of the moving bundle.

Robert Smith^1,2, Daan Christiaens^3,4, Ben Jeurissen⁵, Maximillian Pietsch³, David Vaughan^1,2,6, Graeme Jackson^1,2,6, and J-Donald Tournier^3
1Florey Institute of Neuroscience & Mental Health, Melbourne, Australia, ²Florey Department of Neuroscience and Mental Health, The University of Melbourne, Melbourne, Australia, ³School of Biomedical Engineering & Imaging Sciences, King's College London, London, United Kingdom, ⁴Department of Electrical Engineering, KU Leuven, Leuven, Belgium, ⁵Department of Physics, University of Antwerp, Antwerp, Belgium, ⁶Department of Neurology, Austin Health, Melbourne, Australia

While the Fixel-Based Analysis (FBA) framework provides familywise error control across a whole-brain template accounting for the presence of crossing fibres in the white matter, its typical usage fails to correct for multiple hypothesis tests due to the utilisation of multiple quantitative metrics. We demonstrate different methods that can be employed to provide more comprehensive false positive control in this context.

Jane Ansell^1,2, Irène Brumer³, Jonathan Ashmore⁴, Enrico de Vita⁵, Josef Jarosz², and Marco Borri^2
1King's College London, London, United Kingdom, ²Department of Neuroradiology, King's College Hospital, London, United Kingdom, ³Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, ⁴Department of Medical Physics and Bioengineering, NHS Highland, Inverness, United Kingdom, ⁵Department of Biomedical Engineering, School of Biomedical Engineering and Imagine Sciences, King's College London, London, United Kingdom

Atlas and fMRI-based approaches to standardise seed and end regions for probabilistic tractography of the arcuate fasciculus are investigated. fMRI-based approaches use spheres centered on language task peak activation. Within this pilot cohort, an atlas-based approach demonstrates the greatest sensitivity. fMRI-based approaches are more specific, but sensitivity can be increased by enlarging sphere size. Within each approach, a trade-off between sensitivity and specificity is seen as seed/end region size increases. For patients with abnormalities or lesions, where atlas approaches might be compromised, fMRI methods may be preferred. Further work to optimise fMRI-based approach is warranted, alongside application to patient data.  

Hiromasa Takemura^1,2, Wei Liu³, Hideto Kuribayashi⁴, and Ikuhiro Kida^1,2
1Center for Information and Neural Networks (CiNet), National Institute of Information and Communications Technology, Suita-shi, Japan, ²Graduate School of Frontier Biosciences, Osaka University, Suita-shi, Japan, ³Siemens Shenzhen Magnetic Resonance Ltd, Shenzhen, China, ⁴Siemens Healthcare K.K., Tokyo, Japan

Diffusion MRI measurements of the human optic nerve remain challenging due to susceptibility-induced artifacts, despite their importance in neuro-ophthalmology research. Here, we demonstrate that simultaneous multi-slice (SMS) readout-segmented echo-planar imaging (rsEPI) acquisition has advantages in measuring the human optic nerve over conventional SMS single-shot EPI (ssEPI) with respect to image quality. This result was confirmed with statistical comparisons demonstrating higher fractional anisotropy along the optic nerve in SMS rsEPI data compared with that in SMS ssEPI data. Taken together, this study demonstrates the utility of SMS rsEPI for diffusion MRI measurements of tissue properties of the optic nerve.

Viljami Sairanen^1,2, Mario Ocampo-Pineda¹, Cristina Granziera², Simona Schiavi¹, and Alessandro Daducci^1
1Department of Computer Science, University of Verona, Verona, Italy, ²Translational Imaging in Neurology, Department of Medicine and Biomedical Engineering, University Hospital Basel and University of Basel, Basel, Switzerland

The white matter structures of the human brain can be represented via diffusion tractography. Unfortunately, tractography is prone to find false-positive streamlines causing a severe decline in its specificity and limiting its clinical feasibility. Filtering algorithms have been proposed to reduce these invalid streamlines. We augmented the COMMIT filtering algorithm to adjust for two typical artifacts present in diffusion-weighted images: partial voluming and signal drop-outs due to subject motion. We demonstrate that our robust algorithm is capable to properly filter tractography reconstructions despite these artifacts and could be useful especially for clinical studies with uncooperative patient groups such as neonates.

Erick Hernandez-Gutierrez¹, Ricardo Coronado-Leija², Gabrielle Grenier³, Maxime Descoteaux³, and Alonso Ramirez-Manzanares^1
1Computer Science, CIMAT A.C., Guanajuato, Mexico, ²New York University School of Medicine, New York, NY, United States, ³Université de Sherbrooke, Sherbrooke, QC, Canada

In this work, a novel multi-shell modelling method is proposed to compute fascicle-based fractional anisotropy that is shown accurate and robust in the centrum semiovale of human brain clinical multi-shell datasets. It does not require tractography and it is informed by the local modelling estimations provided by the method called MRDS. The effectiveness of our methodology is also shown on synthetic experiments.

David Lee¹, Ashish Sahib¹, Antoni Kubicki¹, Katherine Narr¹, and Shantanu Joshi^1
1UCLA, Los Angeles, CA, United States

Pairwise inter-tract correlations of the mean diffusion measures have been used to infer structural connectivity among major white matter (WM) fiber tract pathways. Instead of using the mean, we propose a novel approach of constructing kernel density distributions to represent population variability of fractional anisotropy. We also propose the cosine similarity metric between the full shapes of the kernel densities to investigate inter-tract connectivity among the fiber tracts. Our representation and the shape similarity measure is a better predictor of the short- and long-range connectivity disruptions in WM tracts in major depressive disorder patients compared to the conventional mean-based approach.

John Kruper^1,2, Jason D Yeatman^3,4, Adam Richie-Halford², David Bloom^1,2, Mareike Grotheer^5,6, Sendy Caffarra^3,4,7, Greg Kiar⁸, Iliana I. Karipidis⁹, Ethan Roy³, and Ariel Rokem^1,2
1Department of Psychology, University of Washington, Seattle, WA, United States, ²eScience Institute, University of Washington, Seattle, WA, United States, ³Graduate School of Education, Stanford University, Stanford, CA, United States, ⁴Division of Developmental-Behavioral Pediatrics, Stanford University School of Medicine, Stanford, CA, United States, ⁵Center for Mind, Brain and Behavior - CMBB, Hans-Meerwein-Straße 6, Marburg, Germany, ⁶Department of Psychiatry, University of Marburg, Marburg, Germany, ⁷Basque Center on Cognition (BCBL), Brain and Language, Donostia‐San Sebastián, Spain, ⁸Department of Biomedical Engineering, McGill University, Montreal, QC, Canada, ⁹Center for Interdisciplinary Brain Sciences Research, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA, United States

Tractometry from diffusion MRI estimates the tissue properties along the length of major white matter tracts, using: computational tractography; tract segmentation using atlases or other classification methods; and microstructural modeling in voxels along the length of the estimated tracts. Given previous concerns about the sensitivity of dMRI-based analysis to variations in methodology, we tested: the reliability of tractometry results within individuals across measurements; and within measurement, across variations in the analysis methods. We found that although there are variations that arise from differences in tractography methods, bundle segmentation methods, microstructural modeling, and different software implementations, tractometry is overall quite robust.

Alireza Shirazinodeh^1,2 and Hamidreza Saligheh Rad^1,2
1Department of Medical Physics and Biomedical Engineering, Tehran university of Medical Science, Tehran, Iran (Islamic Republic of), ²Quantitative MR Imaging and Spectroscopy Group, Research Center for Molecular and Cellular Imaging, Tehran, Iran (Islamic Republic of)

Tractography of the brain fibers from diffusion signals in magnetic resonance imaging needs to be more sensitive to structures of diffusion images to reduce the number of ODF used to predict tracts and avoid determining false-positive cases in pixels whose values may have changed due to noise or other unwanted factors. The local-based methods can be improved tractography results. In this abstract, we designed a mask based on the local structures of the diffusion phantom images in 64 directions. Using this mask we reduced the number of gradients and false-positive predicted results from the outputs of the tractography algorithm.

Dmitri Shastin^1,2,3, Sila Genc¹, Greg Parker¹, Kristin Koller¹, Chantal M.W. Tax^1,4, John Evans¹, Khalid Hamandi^1,3,5, Derek Jones^1,3, William Gray^1,2,3, and Maxime Chamberland^1
1Cardiff University Brain Research Imaging Centre (CUBRIC), Cardiff University, Cardiff, United Kingdom, ²Department of Neurosurgery, University Hospital of Wales, Cardiff, United Kingdom, ³BRAIN Biomedical Research Unit, Health & Care Research Wales, Cardiff, United Kingdom, ⁴Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands, ⁵Department of Neurology, University Hospital of Wales, Cardiff, United Kingdom

We propose a method of utilising mesh representations of cortical surfaces for generating tractograms of short association fibres facilitating a focused investigation into superficial white matter in health and disease. The method additionally allows to relate streamline metrics to surface topology in native space before registration for comparisons within and between subjects is performed. Our approach is applied to state-of-the-art test-retest data and shows good-to-excellent repeatability.

Yi-Chia Kung¹, Chu-Chung Huang², Chih-Chin Heather Hsu^1,3, Shin Tai Chong¹, Ching-Po Lin¹, and Chun-Hung Yeh^4,5
1Institute of Neuroscience, National Yang-Ming University, Taipei, Taiwan, ²Institute of Cognitive Neuroscience, School of Psychology and Cognitive Science, East China Normal University, Shanghai, China, ³Center for Geriatrics and Gerontology, Taipei Veterans General Hospita, Taipei, Taiwan, ⁴Institute for Radiological Research, Chang Gung University and Chang Gung Memorial Hospital, Taoyuan, Taiwan, ⁵Department of Child and Adolescent Psychiatry, Chang Gung Memorial Hospital, Linkou Medical Center, Taoyuan, Taiwan

This study proposes a novel framework for quantifying the alteration in shape morphometry of white matter (WM) fiber tracts. The surface of each WM bundle is segmented from diffusion MRI data using TractSeg, followed by surface reconstruction, parameterization, and modelling. With the M-rep parameterization and vertex-wise model distance, the proposed method was able  to differentiate the level of morphological lateralization between bilateral arcuate fasciculi and cortical spinal tracts. 

Alexandre Joanisse¹, Guillaume Theaud¹, Jon Haitz Legarreta¹, François Rheault^1,2, and Maxime Descoteaux^1
1Computer Sciences, Université de Sherbrooke, Sherbrooke, QC, Canada, ²Electrical Engineering, Vanderbilt University, Nashville, TN, United States

In most tractography settings, users choose a single tracking algorithm and a single set of tracking parameters, which may limit the quality and impact of the bundle reconstruction results. The present ensemble tractography processing pipeline aims at producing diverse whole brain tractograms faster by tuning and automatically concatenating multiple tractograms from different combinations of parameters and algorithms. Compared to classical tractography, our optimized ensemble tractography obtains improved or similar white matter bundle recovery but in 62% of the processing time.

Ariel Rokem¹, Mauro Bisson², Josh Romero², Thorsten Kurth², Massimiliano Fatica², Pablo Damasceno³, Xihe Xie³, Adam Richie-Halford⁴, Serge Koudoro⁵, and Eleftherios Garyfallidis^5
1Psychology, University of Washington, Seattle, WA, United States, ²NVIDIA, Menlo Park, CA, United States, ³University of California, San Francisco, San Francisco, CA, United States, ⁴University of Washington, Seattle, WA, United States, ⁵Indiana University, Bloomington, IN, United States

Tractography based on diffusion-weighted MRI provides non-invasive in vivo estimates of trajectories of long-range brain connections. These estimates are important in research that measures individual differences in brain connections and in clinical use-cases. But the computational demands of tractography present a barrier to progress. Here, we present a GPU-based tractography implementation that accelerates tractography algorithms implemented as part of the Diffusion Imaging in Python (DIPY) project. This implementation speeds up tractography by at least a factor of ~200X, providing tractographies that closely match CPU-based solutions. These speedups enable applications of tractography in clinical data, and in very large datasets.

Francois Rheault^1,2, Jean-Christophe Houde², Jasmeen Sidhu³, Sami Obaid^2,4, Guido Guberman⁵, Alessandro Daducci⁶, and Maxime Descoteaux^2
1Electrical Engineering, Vanderbilt University, Nashville, TN, United States, ²Computer Sciences, Université de Sherbrooke, Sherbrooke, QC, Canada, ³Faculté de médecine et des sciences de la santé, Université de Sherbrooke, Sherbrooke, QC, Canada, ⁴Centre de Recherche du Centre Hospitalier, Université de Montréal, Montréal, QC, Canada, ⁵Department of Neurology and Neurosurgery, McGill University, Montréal, QC, Canada, ⁶Computer Science, University of Verona, Verona, Italy

Tractography involves complicated processing and connectomics include even more complexity. To facilitate structural connectome reconstruction we present: Connectoflow. Connectoflow requires simple inputs, has simple options and provides simple outputs, all with cutting-edge processing. By leveraging the simplicity of Nextflow and Docker/Singularity, Connectoflow is robust and efficient. By combining Tractoflow with Connectoflow, one can go from raw DW-images to structural connectomes in a few simplified steps. The proposed pipeline innovates by including connection-wise cleaning/filtering, provides connection weights that go beyond streamline count (COMMIT) as well as advanced connection-wise metrics (similarity and AFD).