JOURNÉE SCIENTIFIQUE ANNUELLE - ANNUAL SCIENTIFIC DAY 7 FÉVRIER 2020 - RBIQ
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
JOURNÉE SCIENTIFIQUE ANNUELLE 7 FÉVRIER 2020 ANNUAL SCIENTIFIC DAY FEBRUARY 7, 2020 Images de Luankang Lu, Christopher Steele, Daniel Almeida, et Michael Bernier
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY PROGRAMME / SCHEDULE 8:50 - Bienvenue / Welcome - Martin Lepage and Christine Tardif 9:00 Thème 2 : Bioimagerie fondamentale et modèle animal Thème 1 séance d’affiches : Avancées 9:00 - 3 présentations de 20 min et table ronde méthodologiques 10:30 Theme 2: Basic bio-imaging and animal models Theme 1 poster session: Methodological 3 20-min talks and round table advances 10:30 - Pause café / Coffee break 10:45 10:45 - Nouvelles et annonces du réseau / News and announcements from the network 11:15 Conférence « Étoile montante en bio-imagerie au Québec » / Rising star in bio-imaging in 11:15 - Quebec lecture 12:00 Dr. Maxime Descoteaux, Université de Sherbrooke (Introduction by Christine Tardif) Title TBD 12:00 - Lunch 13:00 Thème 1 : Avancées méthodologiques Thème 3 séance d’affiches : Bioimagerie clinique et 13:00 - 3 présentations de 20 min et table ronde études humaines 14:30 Theme 1: Methodological advances Theme 3 poster session: Clinical bio-imaging and 3 20-min talks and round table human studies 14:30 - Pause café / Coffee break 14:45 Thème 3 : Bioimagerie clinique et études humaines Thème 2 séance d’affiches : Bioimagerie fondamentale 14:45 - 3 présentations de 20 min et table ronde et modèle animal 16:15 Theme 3: Clinical bio-imaging and human studies Theme 2 poster session: Basic bio-imaging and animal 3 20-min talks and round table models Conférence William Feindel / William Feindel lecture 16:15 - Dr. Hedvig Hricak, Cornell University (Introduction by Martin Lepage) 17:15 Oncologic Imaging 2020 and Beyond 17:15 - Cocktail et remise de prix / Cocktail and awards 18:00 7 Février 2020 / February 7, 2020 !2
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY THÈMES / THEMES 1 Avancées méthodologiques / Methodological advances Moderators: Ives Lévesqe, McGill University, & Michèle Desjardins, Université Laval Speakers: Flavie Lavoie-Cardinal, Université Laval Imaging the brain at the nanoscale Jean Provost, Polytechnique Montréal Imaging function with ultrafast ultrasound Jamie Near, McGill University Fun with MRS: developing methods to probe tissue chemistry and metabolism in the brain 2 Bioimagerie fondamentale et modèle animal / Basic bio-imaging and animal models Moderators: Brigitte Guérin, Université de Sherbrooke, & Mallar Chakravarty, McGill University Speakers: Jean Da Silva, Université de Montreal Development of F-18 fluoropyridine analogs of losartan and candesartan for PET imaging of AT1 receptors Brian Nieman, SickKids Brain development after childhood cancer: costs and causes Stephanie Tullo, McGill University Title TBD 3 Bioimagerie clinique et études humaines / Clinical bio-imaging and human studies Moderators: Claudine Gauthier, Concordia University, & Philippe Albouy, Université Laval Speakers: Linda Booij, Concordia University The relevance of peripheral DNA methylation for human brain development and risk for psychopathology Benjamin Morillon, Aix-Marseille University Asymmetric sampling in human auditory cortex reveals spectral processing hierarchy Mathieu Roy, McGill University Neural signatures of pain: can we have objective markers of pain? 7 Février 2020 / February 7, 2020 !3
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY CONFÉRENCIERS INVITÉS / INVITED SPEAKERS Étoile montante en Bio-Imagerie au Québec / Rising Star in Bio-Imaging in Quebec Title of talk TBD Dr. Maxime Descoteaux, Université de Sherbrooke Maxime Descoteaux, PhD, is Professor in Computer Science since 2009 at the Science Faculty of Sherbrooke University. He is the founder and director of the Sherbrooke Connectivity Imaging Laboratory (SCIL) (http://scil.usherbrooke.ca/). His research focuses on brain connectivity from state-of-the-art diffusion MRI acquisition, reconstruction, tractography, processing and visualization. The aim of the SCIL is to better understand structural connectivity, develop novel tractography algorithms, validate them and use them for human brain mapping and connectomics applications. Maxime Descoteaux did a post-doctorat fellow at NeuroSpin under the supervision of Cyril Poupon and Denis Le Bihan. He also obtained a PhD in Computer Science at INRIA Sophia Antipolis - Mediterranée, supervised by R. Deriche after he obtained a M.Sc under the supervision of K. Siddiqi in Computer Science at Center for Intelligent Machines, McGill University, where he also obtained a B.Sc, graduating from the joint honors Mathematics and Computer Science program. Pr Descoteaux holds the USherbrooke Institutional Research Chair in NeuroInformatics. He has been cited more than 6700+ times and has 95+ journal publications, according to google scholar. Conférence William Feindel / William Feindel Lecture: Oncologic Imaging 2020 and Beyond Dr. Hedvig Hricak, Cornell University Hedvig Hricak is Chair, Department of Radiology, Memorial Sloan Kettering Cancer Center, a member of the Molecular Pharmacology and Chemistry Program, Sloan Kettering Institute, and Professor, Gerstner Sloan Kettering Graduate School of Biomedical Sciences, New York, NY, and Professor at Weill Medical College of Cornell University. She is renowned for translating and pioneering applications of novel imaging technologies that address pressing clinical needs, particularly in oncology. She is a member of the National Academy of Medicine and has also been elected a “foreign” member of both the Russian Academy of Sciences and the Croatian Academy of Sciences and Arts. She received honorary doctorates from both Ludwig Maximilian University, Munich, Germany and University of Toulouse III, Paul Sabatier in Toulouse, France. She has served/serves on many academic/government national and international advisory boards. She served on the Board of Scientific Counselors of the National Institutes of Health, the Scientific Advisory Board of the National Cancer Institute, the Advisory Council of the National Institute of Biomedical Imaging and Bioengineering, and the Nuclear and Radiation Studies Board of the National Academy of Sciences. She is a member of the National Cancer Policy Forum of the NAS and presently serves on the External Advisory Board, University of Michigan, Ann Arbor, MI, External Advisory Board of Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD, Advisory Board, University of Vienna, Austria and is a 7 Février 2020 / February 7, 2020 !4
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Member, Scientific Committee of the DKFZ, Germany. Dr Hricak has been president of numerous professional societies, including the Radiological Society of North America (RSNA), the California Academy of Medicine the International Society for Strategic Studies in Radiology (ISSSR) and the Academy for Radiology and Biomedical Imaging. For her many contributions to professional societies and academic , national and government agencies, Dr Hricak received numerous awards including the Marie Curie Award from the Society of Women in Radiology; the gold medals of the International Society for Magnetic Resonance in Medicine, the Association of University Radiologists, the Asian Oceanian Society of Radiology, the European Society of Radiology, and the RSNA; the Béclère medal of the International Society of Radiology; the Schinz Medal of the Swiss Society of Radiology; the Morocco Medal of Merit; the Jean A. Vezina French Canadian Award of Innovation; and the Order of Croatian Morning Star of Katarina Zrinska Presidential Award of Croatia. In 2018, Dr Hricak received the David Rall Medal for Distinguished Leadership from the National Academy of Medicine. 7 Février 2020 / February 7, 2020 !5
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY RÉSUMÉS / ABSTRACTS T1-01 Quatifying white matter disconnection in patient populations Abdelrahman Zayed1, Yasser Iturria-Medina2, Bernhard Sehm3 and Christopher J. Steele4. 1 Department of Electrical and Computer Engineering and PERFORM Centre, Concordia University, Montreal, Quebec, Canada. 2 Department of Neurology and Neurosurgery, McGill University, Montreal, Quebec, Canada. 3 Department of Neurology, Max Planck Institute for Human Cognitive and Brain Sciences and Clinic for Cognitive Neurology, University of Leipzig, Leipzig, Germany. 4 Department of Psychology, Concordia University, Montreal, Quebec, Canada. With an estimated five million new stroke survivors every year and a rapidly aging population suffering from hyperintensities and diseases of presumed vascular origin that affect white matter and contribute to cognitive decline, it is critical that we understand the impact of white matter damage on brain structure and behavior. Current techniques for assessing the impact of lesions/damage consider only location, type, and extent, while ignoring how the affected region was connected to the rest of the brain. Regional brain function is a product of both local structure and its connectivity. Therefore, obtaining a map of white matter disconnection is a crucial step that could help us predict the behavioral deficits that patients exhibit. In the present work, we introduce a new practical method for computing lesion/damage- based white matter disconnection maps that require only moderate computational resources. We achieve this by creating diffusion tractography models of the brains of healthy adults and assessing the connectivity between small regions (nodes). We then interrupt these connectivity models by projecting patients' lesions into them to compute the predicted white matter disconnection. Importantly, the resulting maps of disconnection are patient-specific, quantitative, and therefore enable direct comparison between patients suffering from a variety of lesions in different locations and/or with different behavioral deficits. Our method leverages pre-computed connectivity matrices to reduce storage space and decrease computation time. Each model is reduced from ~6GB (raw streamlines file) to
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY A comparison of locally and publicly implemented QSM pre-processing T1-02 techniques Alex Ensworth1, Véronique Fortier1,2, Ives R. Levesque1,2,3 1 Medical Physics Unit, McGill University 2 Biomedical Engineering, McGill University 3 Research Institute of the McGill University Health Centre Introduction: Annually 2,000,000 cases of Traumatic brain injury (TBI) are reported in North America. Mild TBI is the most common type and currently there is no reliable way of characterizing it. Magnetic Resonance Imaging (MRI) is a promising way of observing brain changes after mild injury, potentially with quantitative susceptibility mapping (QSM). This is complicated due to the common location of the injury, often adjacent to skull bone, which by its magnetic properties degrades the QSM. Our goal is to improve QSM to obtain useful information in the brain cortex adjacent to the skull. Currently, there are many different methods to generate QSMs, but no ground truth. This work demonstrates the comparison of local and public implementation of the various techniques used in the QSM process, in a group of volunteers. The results will inform the pipeline that will be used to quantify mild TBI. Methods: One public and one local implementation of the Laplacian‐based and region growing unwrapping techniques were compared. The highest quality unwrapping technique was then used as the basis for a background removal comparison. One public and one local implementation of the projection onto dipole fields (PDF) and Laplacian Boundary Value (LBV) background removal techniques were compared. The highest quality background removal technique was then used as the basis for the dipole inversion technique. Three different dipole inversion techniques were compared: iterative Susceptibility Weighted Imaging and Mapping (iSWIM), Morphology Enabled Dipole Inversion (MEDI) and Truncated K-space Division (TKD). The unwrapped phase differences and background removed images were compared qualitatively, while the QSMs were compared directly by region of interest extraction and comparing the average values to previously documented susceptibilities in specific brain tissue. Results: The public and local implementations of Laplacian unwrapping disagreed. It is not yet clear which technique agrees with the Quality Guided Region Growing unwrapping. There was no significant difference between implementations for background removal, however the PDF technique proved superior to LBV. The dipole inversion techniques varied in different aspects. Conclusion: While publicly available techniques are available, one should not assume that they function optimally. This work determined that locally implemented techniques were necessary to provide an accurate local field map prior to the dipole inversion step in the QSM pipeline. 7 Février 2020 / February 7, 2020 !7
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Analytical B1 Inhomogeneity Correction for MTsat Values for use in T1-03 Cortical ihMT Mapping Christopher D. Rowley1,2, Zhe Wu1,2,3, Ilana R. Leppert1, Jennifer S.W. Campbell1, David A. Rudko11,2,4, G. Bruce Pike5, Christine L. Tardif1,2,4 1 McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, QC, Canada 2 Department of Neurology and Neurosurgery, McGill University, Montreal, QC, Canada 3 Techna Institute, University Health Network, Toronto, ON, Canada, 4 Department of Biomedical Engineering, McGill University, Montreal, QC, Canada 5 Hotchkiss Brain Institute and Departments of Radiology and Clinical Neuroscience, University of Calgary, Calgary, Canada Introduction: Inhomogeneous magnetization transfer (ihMT) contrast in the brain has been reported to be a more specific myelin biomarker than conventional MT, but can be impacted by B1 inhomogeneities, reducing its accuracy. The ihMT contrast is primarily driven by the size of the semi-solid dipolar coupled pool, where it is produced during single frequency offset (-Δf or +Δf) MT saturation but not in dual frequency MT saturation (-Δf and + Δf). Thus, the difference between dual and single frequency MT-weighted images can provide insight into the dipolar coupled pool. Current methods for calculating ihMT with B1 correction assume a single excitation and readout, either a k-space line or plane, per saturation module. Here we derive equations for use of an arbitrary number of readout segments collected after an arbitrary number of MT pulses in a saturation preparation module. The resulting implementation produces B1 corrected MTsat maps for dual and single frequency MT saturation pulses, which can be combined to generate an ihMT map. Methods: One healthy volunteer was scanned on a 3T Siemens Prisma MRI system. T1 mapping was performed using the variable flip angle method with the following parameters: α1 = 5°, α2 = 20°, TR = 30ms. MT-weighted images were subsequently collected using the following parameters: Δf=7kHz, B1=9.8μT, 12 MT pulses in each preparation module (each RF with 0.96 ms duration, 1.76 ms spacing), α = 5°, 11 readouts per 175ms TR, and 5.3ms echo-spacing for the GRE kernel. The central frequency for the MT pulses was shifted by -100Hz to better align with the center of the lipid pool. All images were collected with 1.5mm isotropic resolution. B1 and B0 maps were also collected. The steady-state signal was solved using Bloch equations to solve for the MTsat values at each voxel using T1, M0, and B1 values to correct the excitation flip angle α, and the MT saturation. ihMTsat was calculated as the difference between the saturation in the dual vs single offset frequency cases. ihMTR was calculated for comparison by using the 5° flip angle image with no MT preparation from the VFA experiment to normalize the signal change. The results were projected onto the mid-cortical surface for analysis. Results and discussion: ihMTR in the cortex displays a strong correlation with B1 (r = 0.46), whereas ihMTsat displays no correlation with B1 (r = 0.07). Both ihMTR (r = 0.74) and ihMTsat (r = 0.75) show correlations with the myelin-sensitive metric R1. Our results support that ihMTsat is a sensitive marker of cortical myelination with minimal bias from B1 inhomogeneity. 7 Février 2020 / February 7, 2020 !8
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Pattern Recognition Algorithms to study Tumour Perfusion using T1-04 Dynamic Contrast-Enhanced MRI Dipal Patel, Zaki Ahmed, Ives Levesque Medical Physics Unit, Faculty of Medicine, McGill University Research Institute of the McGill University Health Centre Introduction: The efficacy of radiotherapy treatments of cancer tumours is related to tumour vasculature, as well- oxygenated tumours are more susceptible to radiation treatment. The irregular vasculature within the tumour results in hypoxic regions with a heterogeneous perfusion distribution. In dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), tumour and tissue perfusion can be described using temporally resolved data to understand contrast agent uptake within the tumour region. This study seeks to identify and adapt a pattern recognition algorithm known as non-negative matrix factorization (NMF) for DCE-MRI to produce quantitative spatial maps of the heterogenous distribution of perfusion in cancer tumors to longitudinally assess treatment response. Methods: The study included patients (n=18) with high-grade soft tissue sarcoma with 3 exams over the course of their neoadjuvant radiotherapy treatments. Imaging was performed serially on a 1.5 T MRI scanner (Signa, GE Healthcare). T1-weighted time-series images were acquired using a 3D Fast Spoiled Gradient Echo sequence: TE=4.2 ms, TR=6.036 ms, flip angle=25 degrees, bandwidth=42 kHz, matrix size=256×128, FOV of 24-28 cm. Each sarcoma was manually contoured by a radiation oncologist and images were masked to exclude healthy tissue from the analysis. Alternating Negative Least Squares using Block Pivot Principle (ANLS-BPP) and Hierarchical Alternating Least Squares (HALS) NMF algorithms were used to identify two (k=2) time-course patterns in the image data and produce the corresponding weight maps. These perfusion curves were normalized and interpreted as high perfusion and low perfusion according to their shape. Perfusion curves and weight maps were compared across timepoints per patient. Results: The NMF algorithms identified source curves that resemble signal enhancement curves for k=2. The source curves for high perfusion and low perfusion patterns show very strong similarities across patients and timepoints. The weight maps that correspond to high and low perfusion curves were superimposed in colour images, revealing the heterogeneity in the perfusion distribution for all tumours, which changes across timepoints. Conclusion: Pattern recognition algorithms can identify perfusion curves in DCE-MRI data and produce the respective weight maps that reflect the heterogenous distribution of perfusion in high grade soft-tissue sarcoma. Pattern recognition could be a more practical and robust data-driven approach to characterizing blood supply in tumours and could be used to study tumour progression through the phases of radiotherapy treatment. 7 Février 2020 / February 7, 2020 !9
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY T1-05 Improving the Speed of Surface Registrations Francis Carter1, Pierre-Louis Bazin2, Christopher Steele1,2 1 Concordia University, Montreal, Quebec, 2 Department of Neurology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany Averaging or comparing brain Magnetic Resonance Imaging (MRI) data between subjects requires image registration. While standard volumetric registration methods typically perform well on subcortical structures, the cortex is much more difficult to co-register due to large amounts of individual variability. Treating the cortex as a 2D manifold to be aligned according to its shape and curvature can result in significantly more accurate registration, but this approach often uses 3D meshes and suffers from greater computational complexity. Previous work by Tardif et al. adapts the computational advantages of volumetric registration and provides equivalent cortical registration accuracy to mesh-based approaches. The goal of the current work was to adapt and implement the main ideas proposed by Tardif et al. to develop a rapid surface-based volumetric registration pipeline in python. The main idea of the method proposed by Tardif et al. was to represent the cortical surface volumetrically as a level set (signed distance function, where values denote the distance to a surface), and then perform registration on the level sets rather than the raw images. This was done using the nighres package and ANTs to develop a python-based pipeline for rapid and efficient volumetric surface-based registration. First, brain segmentation was performed with MGDM in nighres. Second, the CRUISE cortex extraction tool was used to extract a representation of the cortical surface. Third, the extracted contrical boundary level sets were averaged to provide a level set of the middle of the cortex. Lastly, the level sets were defined only within a maximum distance of 10 mm from the cortex, and were used as input to ANTs multivariate template construction to generate the common-space template (3 iterative steps, with default linear and nonlinear registration parameters). Since level sets vary smoothly across the volume, the more computationally efficient Demons image difference metric was used to assess image similarity without affecting accuracy. For standard T1-weighted structural images (1mm3), image segmentation, cortex extraction, and level set fusions take
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Effects of anatomy in a motor and a non-motor region using MRI and T1-06 TMS Francis Houde, Russell Butler, Étienne St-Onge, Marylie Martel, Véronique Thivierge, Maxime Descoteaux, Kevin Whittingstall, Guillaume Léonard. Université de Sherbrooke Introduction: Transcranial magnetic stimulation (TMS) protocols are often performed in non-motor cortical areas. When stimulating non-motor areas with TMS, the most accepted guideline to determine stimulation intensity is the resting motor threshold (rMT; lowest TMS intensity required to elicit a motor evoked potential). This method assumes that all brain regions have the same excitability and homogenous anatomical characteristics so that TMS effects on a region can be directly transposed to another. The aim of this study was to retrospectively assess, from a previous pain memory study, how anatomy influenced the effects of TMS in a motor and in a non-motor brain area. Methods: The effects of TMS on healthy participants (n = 20) over the primary motor cortex (M1; motor region) and over the superior temporal gyrus (STG; non-motor region) were evaluated using the rMT, and pain unpleasantness mnemonic bias, respectively. Pain was evaluated using a 0-10 visual analog scale (VAS) immediately after a painful event and at recall, 2 months later. Pain memory accuracy (VAS recall - VAS initial) was calculated. The attenuation of pain memories was the desired effect (negative scores). Two anatomical variables, both at M1 and STG, were calculated based on TMS stimulations coordinates and anatomical images of the head/brain of the participants (t1-weighted magnetic resonance images): scalp-cortex distance (SCD; mm) and gray matter thickness (GMT; mm). Spearman correlations and Mann Whitney analysis were performed to assess how SCD and GMT affected the rMT and the pain unpleasantness memory bias, in both regions (M1 and STG). Results: SCD at M1, but not STG, correlates with rMT (r = 0.71, p < 0.001). GMT was significantly smaller at M1 when compared to STG (mean M1 = 2.96 ± 0.18, mean STG = 3.27 ± 0.11, p < 0.001) and tended to correlate with rMT at M1 (r = -0.422, p = 0.06), but not STG (p > 0.10). The effect of TMS on STG were more variable (p = 0.048) for participants with STG > M1 SCD (pain mnemonic bias = 0.00 ± 2.67, n = 10) compared to those with M1 > STG SCD (mean pain mnemonic bias = -1.04 ± 1.29, n = 10). GMT at STG tends to correlate with the pain unpleasantness mnemonic bias (r = -0.383, p = 0.096). Conclusion: These observations provide evidences that factors such as SCD and GMT most likely vary across brain regions, as it is the case with M1 and STG in our study. Therefore, we should not assume homogenous anatomical characteristics across regions, and we should strive to develop new approaches, using modeling as example, to account for these anatomical differences when determining which TMS intensity to use in non-motor TMS studies. 7 Février 2020 / February 7, 2020 !11
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY T1-07 The bias of veins on resting state measures of centrality Julia Huck1*, Anna-Thekla Jäger2*, Audrey P. Fan3, Sophia Grahl2, Uta Schneider2, Arno Villringer2,4,5,6, Christine L. Tardif7,8, Pierre-Louis Bazin2,9, Claudine J. Gauthier1,10, Christopher J. Steele2,11 ¹ Concordia University, Department of Physics / PERFORM center, Montreal, Canada, ² Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ³ Stanford University, Stanford, United States, ⁴ Clinic for Cognitive Neurology, University of Leipzig, Leipzig, Germany, ⁵ Leipzig University Medical Centre, IFB Adiposity Diseases, Leipzig, Germany, ⁶ Leipzig University Medical Centre, Collaborative Research Centre 1052-A5, Leipzig, Germany, ⁷ McGill University, Department of Biomedical Engineering, Montreal, Canada, ⁸ Montreal Neurological Institute, Montreal, Canada, ⁹ University of Amsterdam, Faculty of Social and Behavioural Sciences, Amsterdam, Netherlands, ¹⁰ Montreal Heart Institute, Montreal, Canada, ¹¹ Concordia University, Department of Psychology, Montreal, Canada Introduction: Resting-state functional MRI (rs-fMRI) is used to detect low frequency fluctuations using the blood oxygen-level dependent (BOLD) signal. Synchronous fluctuations has been shown to be greater when regions are functionally connected. However, since BOLD is a venous contrast, veins may also introduce a bias in the amplitude and measured location of brain activation¹. These biases are known to affect local BOLD signal and may influence measures of centrality; potentially biasing subsequent inferences regarding connectivity. However, these biases are not well understood. Here, we investigated venous biases on rs-fMRI centrality measures to understand the impact of vein density, diameter and distance on degree centrality (DC) and eigenvector centrality (EC) values². Methods: DC and EC maps were created from the rs-fMRI data of 38 participants with 5 scanning session each using a multiband EPI sequence (TR/TE=1130/22ms; 1.2mm isotropic) on a 7T MRI. Preprocessing included motion correction, fieldmap unwarping, and nuisance regression. DC and EC gray matter maps were computed in native space². Flow-compensated 3D multi-echo gradient images (TR/TE1/TE2=29/8.16/18.35ms; 0.6mm isotropic) were used to reconstruct quantitative susceptibility maps (QSM)³. Veins were segmented from the QSM images⁴. The resulting partial volume (PV) and diameter maps, and the 2.4mm smoothed DC and EC maps were registered to MNI152 space. PV maps were converted to distance maps and the diameter maps were propagated to neighboring regions. Linear regressions were used to identify the relationship between centrality scores and distance in the whole brain and in 7 functional connectivity parcellations⁵. Resulting slopes were tested against the null hypothesis with a two-sided t-test. Vascular density was calculated by dividing the sum of PV voxels within each network by its volume. Results: We found low but significant inverse associations between DC and EC values and distance from all vessel sizes (all p’s
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Design and simulation of a positron detector to measure the AIF for T1-08 dynamic PET Liam Carroll, Shirin A. Enger McGill University Introduction: Dynamic positron emission tomography (dPET) can be used for a more accurate assessment of radiotracer uptake and metabolism than possible with static PET. One promising use for dPET is with gallium-68 (68¬Ga) prostate specific membrane antigen (PSMA) PET for prostate cancer patients where dPET outperforms static PET in identifying cancerous lesions. One factor holding back further research and clinical adoption of techniques, such as the one outlined above, is the need to acquire the arterial input function (AIF) when performing dPET imaging. The purpose of this study was to refine the design and then simulate a previously validated non-invasive positron detector, hereinafter called NID, developed to determine the AIF for dPET. Methods: The NID was simulated using the Geant4 Monte Carlo toolkit. Particle transport was performed by using the Penelope low-energy electromagnetic physics and optical photon processes. The NID consisted of 64 plastic scintillating fibers, 0.97 mm in diameter and 10 cm long. Each end of the fibers was coupled to a photomultiplier tube (PMT). The fibers were arranged around a 64.13 mm diameter polyethylene cylinder that represents a patient’s wrist. Two 2.30 mm diameter cylinders were placed inside the wrist to simulate the radial artery and vein of a patient. These cylinders were 6 mm apart and 2 mm below the surface of the wrist. Simulations were performed by simulating decay events of either oxygen-15 (15O) or fluorine-18 (18F) where the decaying particles were randomly distributed in the artery and in the vein. Interactions were recorded when energy was deposited in the scintillators. Total deposited energy in the scintillator was calculated and the location of interaction for each event was scored. The number of optical photons that accumulated in each PMT were also calculated. The data was analyzed using a python algorithm to separate the arterial signal from the venous signal. The location of interaction was determined by taking the PMT that had the highest signal per event. Results: With 15O, the arterial signal produced a full width half max (FWHM) of 5.28 mm in the cross-section between the artery and the vein. The arterial signal was 104% of the true arterial signal. With 18F, the arterial signal produced a FWHM of 10.32 mm. The arterial signal was 112% of the true arterial signal. The algorithm was able to correctly determine the location of interaction in the scintillator with an accuracy of 98%. Conclusion: The results show that the NID has sufficient spatial resolution to distinguish between the radial artery and radial vein in a patient and is thus suitable to determine the AIF during dPET scans. 7 Février 2020 / February 7, 2020 !13
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Shear wave elastography potential to characterize tissues in breast T1-09 cancer during radiotherapy Marie-Claude Lehoux1,2, Ivan Dimov1,2,3, Maria Bichay1,2,4, François Vincent1,2, Frédéric Chapuis1,2, François Berthod5,6, Guy Cloutier7,8,9,10, Antony Bertrand-Grenier1,2,11 1 Centre intégré universitaire de santé et de services sociaux de la Mauricie-et-du-Centre-du-Québec (CIUSSS MCQ), Trois-Rivières, Québec, Canada. 2 Centre hospitalier affilié universitaire régional, CIUSSS MCQ, Trois-Rivières, Québec, Canada. 3 Faculté de médecine, Université de Montréal, Montréal, Québec, Canada. 4 Département de génie mécanique, Université du Québec à Trois-Rivières, Trois-Rivières, Québec, Canada. 5 Laboratoire en organogénèse expérimentale (LOEX), Centre de recherche du Centre hospitalier universitaire de Québec (CHUQ), Université Laval, Québec, Québec, Canada. 6 Département de chirurgie, faculté de médecine, Université Laval, Québec, Québec, Canada. 7 Centre de Recherche du Centre Hospitalier de l’Université de Montréal (CRCHUM), Montréal, Québec, Canada. 8 Laboratoire de biorhéologie et d’ultrasonographie médicale (LBUM), CRCHUM, Montréal, Québec, Canada. 9 Département de radiologie, radio-oncologie et médecine nucléaire, Université de Montréal, Montréal, Québec, Canada. 10 Institut de génie biomédical, Université de Montréal, Montréal, Québec, Canada. 11 Département de chimie, biochimie et physique, Université du Québec à Trois-Rivières, Trois-Rivières, Québec, Canada. Introduction : La lutte contre le cancer est souvent entravée par les résistances physiques des tumeurs et les dommages collatéraux que causent les traitements. En plus d'être un marqueur de malignité, la rigidification tumorale peut promouvoir la prolifération des cellules cancéreuses et la migration des métastases. La radiothérapie est un traitement utilisant la radiation pour détruire les cellules cancéreuses. La difficulté de la radiothérapie réside dans la destruction de ces cellules tumorales tout en épargnant les tissus sains environnants. Évaluer la réponse au traitement de la tumeur est un enjeu très important afin d’éviter le décès du patient et diminuer les risques de récidive. L’élastographie dynamique ultrasonore (SWE) permet de mesurer des propriétés mécaniques des tissus. Cette technique pourrait permettre d’observer l’évolution des propriétés mécaniques de la zone tumorale et des tissus tout au long du traitement de radiothérapie. Méthodes : 10 patientes ont été recrutées afin d’avoir 5 examens de SWE, soient avant (1), pendant (3) et après (1) les traitements. Des corrélations des mesures de SWE ont été effectuées notamment avec la paroi tumorale, l’épaisseur de peau, la dose et le temps. Résultats : Le diamètre et l’aire de la cavité et la paroi tumorale ont diminué de façon significative (P < 0,05). Les valeurs de SWE (vitesse d’ondes de cisaillement) fut statistiquement significative pour la quasi-totalité des différentes régions entre-elles (lesion, sein affecté, côté symétrique, sein non affecté) (P < 0,05). Trois patientes ont eu un cancer plusieurs années auparavant et des différentes statistiques significatives ont été observés en comparant celles-ci aux autres patientes n’ayant pas eu de cancer auparavant (P < 0,001). Des différences significatives en ce qui a trait au temps a été observés pour les valeurs de SWE de la lésion (p = 0,026; (rs) = -0,693) et du sein affecté (p = 0,036; (rs) = -0,841), alors qu’aucune différence significative a été observé pour le sein « contrôle ». L’épaisseur de peau du sein affecté (1,99±0,38 mm) a été significativement différente du sein contrôle (1,48±0,29 mm). Conclusion : Ce projet présente l’application du SWE au suivi de traitement du cancer du sein par radiothérapie. L’étude des propriétés mécaniques des tissus permettrait d’obtenir de nouvelles informations sur la radiorésistance des tumeurs et la réponse au traitement. Les données obtenues grâce au SWE pourraient être utilisées afin d’adapter les planifications de radiothérapie pour chaque patiente dans le but d’optimiser les traitements, diminuer le taux de mortalité, diminuer les risques de récidive et épargner les tissus sains. 7 Février 2020 / February 7, 2020 !14
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Building a cortical brain MRI atlas following a phylogenetic approach: T1-10 first quantitative results Maryna Zhernovaia1, Mahsa Dadar2,3, Yashar Zeighami3, Josefina Maranzano1,2,3 1 Université du Québec à Trois-Rivières 2 Université Laval 3 McGill University Background: Several cortical atlases have been developed to perform a consistent division of the human cortex in areas that have common structural as well as meaningful and distinctive functional characteristics. Our study proposes the division of cortical areas following a phylogenetic approach. Objective: to construct an MRI atlas that divides the cortex into five main regions of interest (ROIs): 1-archicortex (AC), 2-paleocortex (PC), 3-peri-archicortex (PAC) 4- proisocortex (PIC), 5-temporopolar-isocortex (TPIC). Methods: MRI scans: 1) T1weighted (T1w) average MNI ICBM 152 non-linear 6th generation symmetric average brain MRI model; 2) T1w and magnetization transfer ratio (MTR) images of 10 healthy participants (HP) from the Cambridge Centre for Ageing and Neuroscience data repository. Image analysis: Manual segmentation on the MNI ICBM model of the five ROIs using the interactive software package ‘Display’, which allows the creation of cortical masks for each ROI. The native T1w of each HP was registered to the model using a linear transformation. A hierarchical multiscale non-linear fitting algorithm technique was used to obtain the deformation vector field that maps the HP T1w to the template, and the inverse of the non-linear HP-to-template transformation was used to obtain the native T1w cortical masks of the ROIs. The masks obtained were manually corrected. Dice-Kappa were calculated between the automated and corrected masks to quantify the manual corrections per ROI. Volumes and MTR values of each ROI were computed and compared. Results: Volumes of left AC, PC, PAC, PIC were not significantly different than the right volumes. The right TPIC was significantly larger than the left one: 3.9cc vs 3.3cc (t=3.4, p
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY TRAMPOLINO: the Swiss Army Knife for exploratory tractography and T1-11 reproducible workflows Matteo Mancini1,2,3, Tommy Boshkovski2, Agah Karakuzu2, Elizabeth Dupre4, Jean-Baptiste Poline4, Bruce Pike5, Jennifer Campbell4, Mara Cercignani1, Nikola Stikov2 1 Department of Neuroscience, Brighton and Sussex Medical School, University of Sussex, Brighton, UK 2 NeuroPoly Lab, Polytechnique Montreal, Montreal, Canada 3 CUBRIC, Cardiff University, Cardiff, UK 4 Montreal Neurological Institute, McGill University, Montreal, Canada 5 Hotchkiss Brain Institute and Department of Radiology, University of Calgary, Calgary, Canada Introduction: Across the neuroimaging field, several tools have been developed to process diffusion MRI data, in particular for tractography. Despite all of the promising efforts, the unavoidable presence of false positives and negatives still makes comparison different methods and understanding of parameter sensitivity necessary to improve tractography. Here we introduce TRAMPOLINO (TRActography Meta-Pipeline cOmmand LINe tOol), a tool based on Nipype that aims to provide an immediate command-line interface to the most popular software packages available. The goal of this tool is to offer a framework for reproducible workflows with a common interface to the already implemented tractography approaches. Methods: TRAMPOLINO defines and organizes workflows in a modular fashion, focusing on the three main steps: reconstruction, where the diffusion data are used to estimate either a tensor or a fiber orientation distribution (FOD) for each voxel; tracking, where the streamlines are reconstructed on the basis of the FOD/tensor; and filtering, where using selected constraints or assumptions are used to identify and remove spurious. As a consequence of its modular architecture, it is possible to run all the three steps or any subset of them. Each workflow is specific to one of the processing steps and consists of a set of nodes interconnected with each other. Each node represents a Nipype interface to a specific command-line tool and has specific input and output fields. The workflows are implemented using a common input/output structure for each processing step: in this way, it is possible to easily compare different workflows when using common parameters. The tracking step offers tailored features to make exploratory tractography easy and immediate. Regardless of the specific workflow, it facilitates automatic generation of multiple outcomes, providing a comma-separated list of available algorithms, angular threshold and minimal length constraints. For the angular threshold in particular it is possible to generate several results for a range of thresholds defined by the lower and upper limits. Conclusions: Tractography is an ever-growing field and the availability of different methods and implementations has made clear its tremendous potential but also its current limits. For the sake of both research and clinical applications, reproducibility and variability across methods become fundamental topics to explore, and it is therefore necessary to have a clear and shared interface with the software that has been developed so far. With TRAMPOLINO, we want to provide such a tool. 7 Février 2020 / February 7, 2020 !16
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY The Effect of Noise Correction Techniques on Resting-State Functional T1-12 Connectivity Michalis Kassinopoulos1, Georgios D. Mitsis2 1 Graduate Program in Biological and Biomedical Engineering, McGill University 2 Department of Bioengineering, McGill University It is well established that confounding factors related to head motion and physiological processes (e.g. cardiac and breathing activity) should be taken into consideration when analyzing and interpreting results in fMRI studies. However, even though recent studies aimed to evaluate the performance of different preprocessing pipelines there is still no consensus on the optimal strategy. This may be partly because the quality control (QC) metrics used to evaluate differences in performance across pipelines often yielded contradictory results. Importantly, noise correction techniques based on physiological recordings or expansions of tissue-based techniques such as aCompCor have not received enough attention. Here, to address the aforementioned issues, we evaluate the performance of a large range of pipelines by using previously proposed and novel quality control (QC) metrics. Specifically, we examine the effect of three commonly used practices: 1) Removal of nuisance regressors from fMRI data, 2) discarding motion-contaminated volumes (i.e., scrubbing) before regression, and 3) low-pass filtering the data and the nuisance regressors before their removal. To this end, we propose a framework that summarizes the scores from eight QC metrics to a reduced set of two QC metrics that reflect the signal-to-noise ratio (SNR) and the reduction in motion artifacts and biases in the preprocessed fMRI data. Using resting-state fMRI data from the Human Connectome Project, we show that the best data quality, is achieved when the global signal (GS) and about 17% of principal components from white matter (WM) are removed from the data. In addition, while scrubbing does not yield any further improvement, low-pass filtering at 0.20 Hz leads to a small improvement. 7 Février 2020 / February 7, 2020 !17
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Quantitative imaging performance of MARS spectral photon-counting T1-13 CT for radiotherapy Mikaël Simard, Raj Panta, Stephen Bell, Anthony Butler, Hugo Bouchard 1 Département de Physique, Université de Montréal 2 School of Physical and Chemical Sciences, University of Canterbury, Christchurch, New Zealand Purpose: To evaluate the quantitative imaging performance of novel CT technology, a spectral photon-counting CT (SPCCT), for radiotherapy applications. Specifically, an experimental comparison of the quantitative performance of a clinical Siemens dual-energy CT (DECT) and a Medipix All Resolution System (MARS) SPCCT is performed to estimate physical properties relevant to radiotherapy of contrast agent solutions and human substitute materials. In human substitute materials, the accuracy of quantities relevant to photon therapy, proton therapy and Monte-Carlo simulations, such as the electron density, proton stopping power and elemental mass fractions is evaluated. For contrast agent solutions, the accuracy of the contrast agent concentrations as well as the virtual non-contrast (VNC) electron density is evaluated. Methods: Human tissue substitute phantoms (Gammex 467 and 472) as well as diluted solutions of contrast agents (iodine and gadolinium based) are scanned with commercial systems : a Siemens SOMATOM Definition Flash dual- source CT and a MARS spectral photon-counting micro-CT (MARS V5.2, MARS Bioimaging Ltd., Christchurch, New Zealand). Material decomposition is performed in a maximum a posteriori framework with an optimized material basis tailored to characterize human substitute materials and contrast agents in the context of experimental multi-energy CT data. Results: The root-mean-square error (RMSE) of the electron density calculated over all Gammex plugs is reduced from 1.09 to 0.89% when going from DECT to SPCCT. For the proton stopping power, the RMSE is reduced from 1.92 to 0.89%. Elemental mass fractions of hydrogen, carbon, nitrogen, oxygen and calcium are more accurately estimated with the MARS. The RMSE on the iodine-based contrast agents concentration is reduced from 0.27 to 0.12 mg/mL with SPCCT, and the VNC electron density from 0.40 to 0.22%. A reduction of beam hardening artifacts is also observed on parametric maps of iodine concentration with the MARS. Conclusion: In the present phantom study, a MARS photon-counting scanner provides superior accuracy than a Siemens SOMATOM Definition Flash DECT scanner to quantify physical parameters relevant to radiotherapy. This work experimentally demonstrates the benefits of using more energies in computed tomography to characterize human tissue equivalent materials. This highlights the potential of SPCCT for particle therapy, where more accurate tissue characterization is needed, as well as for Monte-Carlo based planning, which requires accurate elemental mass fractions. 7 Février 2020 / February 7, 2020 !18
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY T1-14 The Modular Organization of Heritability Across the Cortex Nadia Blostein1,7, Sejal Patel1,4,5, Gabriel A. Devenyi1,6, Raihaan Patel1,2, M. Mallar Chakravarty1,2,3,6 1 Cerebral Imaging Centre, Douglas Mental Health University Institute, Verdun, Canada 2 Department of Biological and Biomedical Engineering, McGill University, Montreal, Canada 3 Integrated Program in Neuroscience, McGill University, Montreal, Canada 4 Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Canada 5 Institute of Medical Science, University of Toronto, Toronto, Canada 6 Department of Psychiatry, McGill University, Montreal, Canada 7 Faculty of Arts & Science, McGill University, Montreal, Canada Introduction: Heritability estimates are a straightforward gauge of the specificity of additive genetic effects and can be extended to a bivariate model in order to examine the genetic relationship between two phenotypes. It has been shown that cortical thickness (CT) and surface area (SA) across regions of interest (ROIs) are moderately to highly heritable. Given the connectomic architecture of the brain, it is unlikely that traits such as CT and SA are inherited without any dependence on more distal brain regions. The current project extends previous work from our group to examine the shared heritability and genetic cross-correlation of CT and SA across the human cortex, using a twin and non-twin sibling heritability design. Methods: We obtained 3 Tesla T1-weighted structural magnetic resonance images of 875 healthy adult twins and non- twin siblings (Human Connectome Project), using the CIVET/2.1.0 pipeline. CIVET averaged vertex-wise CT and summed vertex-wise SA within 78 cortical ROIs, predefined by the Automated Anatomical Labeling atlas. Manual quality control for image segmentation and removal of subjects without siblings reduced the sample size to 757. The OpenMx package (2.12.2) in R (3.5.1) was used to compute the shared heritability and genetic cross-correlation (degree of genetic overlap between two traits) of the CT and SA between each possible pair of cortical ROIs. A structural correlation matrix of the CT and SA between each pairwise combination of ROIs was computed. Each of these three matrices were hierarchically clustered using the Python 3.6.1 sklearn 0.18.1 package. This segregated the cortical regions into modules that were either mediated by the same genetic factors or highly structurally correlated. Results: The heritability model did not output significant results for CT. The shared heritability of the SA between pairwise combinations of cortical ROIs is significant (p < 0.05) and does not form clusters. The genetic cross-correlation of the SA between pairwise combinations of cortical ROIs is significant (p < 0.05) and forms four modules. The structural correlation matrices of ROI-wise CT and SA cluster into four modules that do not overlap with the genetically correlated clusters. Conclusion: These results show that there are four cortical SA modules that are mediated by the same genetic factors and that they are not driven by structural correlation. Given the spatially heterogeneous laminar structure of the cortex, gene expression plays a significant in corticogenesis. These cortical modules could therefore inform future studies aiming to parse the relationship between neurodevelopment and human-specific cortical expansion. 7 Février 2020 / February 7, 2020 !19
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Baseline hippocampal grading predicts cognitive decline in subjects T1-15 with mild Alzheimer’s disease Neda Shafiee, Mahsa Dadar, D. Louis Collins McConnell Brain Imaging Center, Montreal Neurological Institute, McGill University, Montreal, Canada Early detection of Alzheimer’s disease (AD) enables early therapeutic intervention which may increase the effectiveness of treatment. The hippocampus (HC) has shown early involvement in the pathological process of AD and can be measured in vivo with MRI. Consequently, abnormal patterns of atrophy in the HC could be used as an early biomarker in the progression of AD [Frisoni et al., 2010]. In order to quantify AD-related atrophy patterns, we used the Nonlocal Image Patch Estimator (SNIPE) method [Coupé et al., 2012]. We then used the SNIPE score to predict cognitive decline from a single visit in individuals with mild AD. Data were obtained from The Alzheimer’s Disease Neuroimaging Initiative (ADNI) database including ADNI1, ADNI2, and ADNIGO. The mild AD cohort (N=100) was chosen based on the Clinical Dementia Rating (CDR=0.5) and the Mini-Mental State Exam (MMSE, range 24-30) scores. This cohort was then divided into two subgroups based on the changes in CDR score in the two years follow up clinical visits. Subjects with more than one point increase in CDR were considered having cognitive decline (N=41). The HC was first segmented using a non-local patch-based method [Coupé et al., 2011]. We then computed SNIPE scores for left and right HC using a template library image drawn from the ADNI1 dataset. SNIPE aims to learn the non-local similarity between the target subject and a training set including both healthy controls and AD cohort data and assigns a similarity score to each voxel in the target structure. We trained a decision tree classifier to predict cognitive decline in the next two years using a feature set including HC SNIPE scores as the MRI biomarkers, sex, and age. The classifier was then tested using leave one out cross-validation. Results showed that there is a significant difference between the two groups in hippocampus SNIPE scores (
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY Décoder le locus spatial de l’attention visuelle volonraire à partir des T1-16 ERPs Pénélope Pelland-Goulet, Martin Arguin, Pierre Jolicoeur Université de Montréal La composante N2pc est un indicateur du déploiement spatial latéralisé de l’attention. La SPCN, quant à elle, est une composante associée à la mémoire de travail. L’étude propose d’utiliser ces deux marqueurs électrophysiologiques (ÉEG) afin de suivre le décours temporel et spatial du déploiement de l’attention à travers une étendue horizontale où sont disposées des lettres. Dans une tâche de Posner modifiée, les essais consistaient en l’apparition d’une flèche centrale indiquant la position vers laquelle déployer son attention, suivie, 750ms plus tard, de l’apparition de lettres aléatoires dont le moment d’apparition à travers une période de 200 ms était aléatoire (durée de 33ms). Après une pause de 2 secondes, les participants avaient comme instruction d’inscrire au clavier la lettre présentée à la position indicée. Le signal ÉEG associé au déploiement visuo-spatial de l’attention pour chaque participant a ensuite été fourni à un Support Vector Machine, qui a dans un premier temps classifié la position à laquelle la lettre indicée était positionnée, puis, dans un deuxième temps, classifié la présence vs l’absence d’attention à une localisation particulière. Le locus attentionnel a été prédit correctement dans 51,7% des cas (niveau de chance à 25%; p < 0,01) et la présence vs l’absence d’attention dans 75% des cas (niveau de chance à 50%, p < 0,001). Il est intéressant de noter que, prise individuellement, la composante SPCN offre une meilleure précision de décodage que la N2pc, suggérant un apport significatif d’information quant à la position spatiale de la lettre traitée. 7 Février 2020 / February 7, 2020 !21
JOURNÉE SCIENTIFIQUE ANNUELLE DU RBIQ / QBIN ANNUAL SCIENTIFIC DAY The R1- weighted connectome: complementing brain networks with a T1-17 myelin-sensitive measure Tommy Boshkovski1, Ljupco Kocarev2, Julien Cohen-Adad1,3,4, Bratislav Misic5, Stéphane Lehéricy6 Nikola Stikov1,7 Matteo Mancini1,8,9 1 NeuroPoly Lab, Polytechnique Montreal, Montreal, Canada 2 Macedonian Academy of Sciences and Arts, Skopje, Macedonia 3 Department of Neurosciences, Faculty of Medicine, University of Montreal, QC, Canada 4 Functional Neuroimaging Unit, Centre de recherche de l’institut universitaire de gériatrie de Montréal 5 Montreal Neurological Institute, Montreal, QC, Canada 6 Institute for brain and spinal cord (ICM), Centre for NeuroImaging Research, Paris, France 7 Montreal Heart Institute, Montreal, QC, Canada 8 Department of Neuroscience, Brighton and Sussex Medical School, University of Sussex, Brighton, United Kingdom 9 CUBRIC, Cardiff University, Cardiff, United Kingdom Introduction: Myelin is one of the key factors that determines how fast signals travel through white matter pathway. Several studies have used myelin-sensitive MR measures in brain network models. Among those measures, the longitudinal relaxation rate (R1) has not been sufficiently explored in connectomics. In this work, we characterized networks weighted by R1 in comparison with the use of number of streamlines (NOS) as a weight. Methods: 10 healthy subjects (HC) (7F/3M,mean age±sd:69.5±9.5) participated in the present study. Each participant was scanned using the following acquisition protocol on a 3T SIEMENS Prisma: diffusion-weighted imaging (DWI) quantitative R1 relaxation rate. The brain was extracted from the MP2RAGE images using the BET toolbox from FSL. The skull-stripped UNI image was processed using FreeSurfer 6.0 to segment the different tissues and parcellate the brain using the Desikan-Killiany Atlas. To avoid bias from the different size of the parcels, we subdivided them into finer regions with the same size using the Lausanne 2008 parcellation (scale 125). The diffusion images for each subject were corrected for motion and eddy current distortions and registered to the skull- stripped UNI image using affine registration. Using a deterministic tractography algorithm, we reconstructed the tractogram and the connectivity matrix was assembled using two different weights: NOS and associated R1. Group consensus networks, for both NOS- and R1-weighted connectomes, were created by taking the median across the subjects. Then, we computed the strength distribution and the weighted average of respectively the NOS-weighted and R1-weighted connectomes. We then compared the community structure of the group connectomes using Louvain algorithm. To further explore the modular structure, a gradient analysis was performed comparing the two connectomes. Results: The results showed that the weighted average distribution of the R1-weighted connectome differs from the strength distribution of the NOS-weighted connectome. The weighted average distribution of the R1-weighted connectome does not show heavy-tailed distribution. Both connectomes exhibited similar community structure. However, the gradient analysis showed that highly myelinated subnetworks (eg. primary motor cortex) do not necessarily coincide with highly connected ones. Conclusion: In summary, we showed that R1-weighted connectomes provide a different perspective into the white matter organization. As future work, we plan to conduct a more comprehensive analysis on a bigger dataset to use the R1-weighted connectome in functional and pathological applications. 7 Février 2020 / February 7, 2020 !22
You can also read