Luke Joel Edwards¹, Kerrin J. Pine¹, Shubhajit Paul¹, Fakhereh Movahedian Attar¹, Michael Herbst², Mirsad Mahmutović³, Boris Keil³, Harald Möller⁴, Evgeniya Kirilina^1,5, and Nikolaus Weiskopf^1,6
1Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ²Gengenbach, Germany, ³Institute of Medical Physics and Radiation Protection, TH Mittelhessen University of Applied Sciences, Giessen, Germany, ⁴NMR Group, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ⁵Center for Cognitive Neuroscience Berlin, Freie Universität Berlin, Berlin, Germany, ⁶Felix Bloch Institute for Solid State Physics, Leipzig University, Leipzig, Germany

Spiral diffusion weighted imaging (DWI) with field monitoring and iterative reconstruction offers potentially reduced echo time (TE) and higher effective resolution (less blurring) compared to EPI. Coupled with a scanner with ultra-strong gradients, it enabled an 800 µm DWI protocol for imaging fine structures of the brain in vivo. Compared to EPI, the shorter TEs provided distinctly different contrast in iron-rich areas (U-fibres and sub-cortical nuclei), which could enhance investigations of these regions. The protocol did, however, come with a reduction in SNR/(unit time) compared to EPI due to differences in readout time.

Yuxi Pang^1
1Dept. of Radiology, University of Michigan, Ann Arbor, MI, United States

Orientation-dependent transverse ($$$R_2$$$ and $$$R_2^*$$$) relaxation phenomena have been documented in the human brain white matter, yet the underlying relaxation mechanisms still remain not well understood. This work is to propose an alternative relaxation pathway through restricted molecular rotational diffusion, in terms of a generalized magic angle effect (gMAE) model, to better characterize recently reported anisotropic $$$R_2$$$ of myelin water and intra- and extracellular water in vivo at 3T. The proposed gMAE model is intrinsically connected with anisotropic translational diffusion from DTI, elucidating not only previously reported $$$R_2^*$$$ anisotropy ex vivo but also its temperature-dependence at 7T.

Cornelius Eichner¹, Michael Paquette¹, Guillermo Gallardo¹, Christian Bock², Jenny E. Jaffe^3,4, Carsten Jäger¹, Evgeniya Kirilina^1,5, Ilona Lipp¹, Toralf Mildner¹, Torsten Schlumm¹, Felizitas C Wermter², Harald E. Möller¹, Nikolaus Weiskopf¹, Catherine Crockford^4,6, Roman Wittig^4,6, Angela D Friederici¹, and Alfred Anwander^1
1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ²Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany, ³Project Group Epidemiology of Highly Pathogenic Microorganisms, Robert Koch Institute, Berlin, Germany, ⁴Tai Chimpanzee Project, Centre Suisse de Recherches Scientifiques en Cote d'IVoire, Abidjan, Cote D'ivoire, ⁵Center for Cognitive Neuroscience Berlin, Freie Universität Berlin, Berlin, Germany, ⁶Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

Detailed neuroanatomical comparisons between humans and chimpanzees could greatly benefit evolutionary neuroscience. However, ethical considerations regarding primate research disallow acquisitions of chimpanzee MRI data in vivo for multiple years. Hence, the availability of diffusion MRI (dMRI) and tractography data from chimpanzees is limited to a few previously acquired datasets. Here we optimize diffusion acquisitions for an interdisciplinary approach to great ape neuroimaging, using post-mortem dMRI data from naturally deceased wild and captive animals. The optimization of data quality from two acquisition strategies allowed to us acquire chimpanzee diffusion MRI data of unpreceded quality and reopen a gateway for evolutionary neuroscience.

Wonil Lee¹, Byungjai Kim¹, Jongyeon Lee¹, and HyunWook Park^1
1KAIST, Daejeon, Korea, Republic of

Many studies have been performed to show that IVIM could be used as a biomarker for various diseases(1-5). Since IVIM is formulated by a biexponential model, it is difficult to quantify the IVIM parameters. Researchers have tried to solve the inverse problems of the biexponential model using two approaches:improving fitting method and selecting optimized b-values(6-8). The trained DNN and the optimized b-values by the proposed method quantified IVIM parameters more accurately than combination of the conventional b-value optimization schemes with DNN fitting method. The optimized b-values by the proposed method showed superior performance even when combined with other fitting methods.

Liam S. P. Lawrence¹, Rachel W. Chan², Hanbo Chen³, Brian Keller³, James Stewart³, Mark Ruschin³, Brige Chugh^3,4, Mikki Campbell³, Aimee Theriault³, Greg J. Stanisz^1,2,5, Scott MacKenzie³, Sten Myrehaug³, Jay Detsky³, Pejman J. Maralani⁶, Chia-Lin Tseng³, Greg J. Czarnota^1,2,3, Arjun Sahgal³, and Angus Z. Lau^1,2
1Medical Biophysics, University of Toronto, Toronto, ON, Canada, ²Physical Sciences, Sunnybrook Research Institute, Toronto, ON, Canada, ³Department of Radiation Oncology, Sunnybrook Health Sciences Centre, Toronto, ON, Canada, ⁴Department of Physics, Ryerson University, Toronto, ON, Canada, ⁵Department of Neurosurgery and Paediatric Neurosurgery, Medical University of Lublin, Lublin, Poland, ⁶Department of Medical Imaging, University of Toronto, Sunnybrook Health Sciences Centre, Toronto, ON, Canada

Technical performance evaluation of diffusion parameters on MR-Linacs (MRLs) is important for cancer applications. We evaluated the accuracy and repeatability of 1.5T MRL measurements of apparent diffusion coefficient (ADC) and intravoxel incoherent motion blood volume fraction (IVIM-f) in the brain via comparison with a diagnostic-quality scanner, in patients undergoing treatment. ADC measurements agree in normal and tumour tissue, but are biased in cerebrospinal fluid. IVIM-f measurements are likely negatively biased. Repeatability is comparable between scanners. The majority of high-grade glioma patients demonstrated significant ADC changes, but not IVIM-f changes.

Maëlig Chauvel¹, Ivy Uszynski¹, William Hopkins², Jean-François Mangin¹, and Cyril Poupon^1
1Université Paris-Saclay, CEA, CNRS, BAOBAB, Neurospin, Gif-sur-Yvette, France, ²Keeling Center for Comparative Medicine and Research, The University of Texas MD Anderson Cancer Center, Bastrop, TX, United States

Mapping the chimpanzee brain connectome is a key to the comprehension of the singularity of the human brain evolution. Contrary to the macaque species, few studies have been performed in vivo on chimpanzees due to its proximity with humans. While different atlases of the structural connectivity exist for humans, we established the first superficial atlas of the white matter connectivity in chimpanzees, using diffusion MRI-based tractography and advanced fiber clustering techniques.

Jesus E. Fajardo¹ and Gonzalo A. Álvarez^1,2,3
1Centro Atómico Bariloche, CONICET, CNEA, 8400, San Carlos de Bariloche, Argentina, ²Instituto Balseiro, CNEA, Universidad Nacional de Cuyo, 8400, San Carlos de Bariloche, Argentina, ³Instituto de Nanociencia y Nanotecnología, CONICET, CNEA, San Carlos de Bariloche, Argentina

Detecting non-invasively tissue-microstructural changes associated with pathologies constitutes a promising diagnosis paradigm. Susceptibility-induced magnetic field gradients constitute a way to obtain microstructural information which is expected to complement Diffusion Weighted Imaging (DWI) methods, since these methods arise from distinct physical phenomena. We performed simulations that show the potential of extracting quantitative tissue-microstructure information of axons in the white matter based on internal-gradient-distribution tensors. We show promising results that can allow detecting alterations in axon diameters, axon density and demyelination using internal gradient distributions as a diagnostic tool.

Stephanie Allan Crater¹ and Nian Wang^2
1Duke University, Durham, NC, United States, ²Radiology and Imaging Sciences, Indiana University, Indianapolis, IN, United States

Tractography and structural connectome have been widely used in preclinical studies in recent years. However, assessment of the experimental parameters for the final outputs of diffusion MRI is still lack. In this study, we acquired an ex vivo mouse brain high spatial (50 µm isotropic) and angular resolution (384 diffusion encoding directions) diffusion MRI dataset in a preclinical 9.4T system. We analyzed the tractography and connectome under different experimental conditions, including b value, spatial resolution, and angular resolution.

Kulam Najmudeen Magdoom¹, Alexandru V. Avram¹, Dario Gasbarra², Qiuyun Fan³, Thomas Witzel³, Susie Y Huang³, and Peter J Basser^1
1National Institute of Health, Bethesda, MD, United States, ²University of Helsinki, Helsinki, Finland, ³Massachusetts General Hospital and Harvard Medical School, Charlestown, MA, United States

Measuring and mapping the diffusion tensor distribution (DTD) via MRI holds the promise of revealing the tissue microstructure at sub-voxel resolution. In this study, we show the DTD results obtained on a human brain in-vivo using the new framework at MGH Connectome scanner with 300 mT/m gradients. The DTD within a voxel is assumed to be a normal tensor variate distribution constrained (CNTVD) within the manifold of 3 x 3 symmetric positive definite matrices. The estimated DTD is used to obtain a family of metrics which aim to disentangle size, shape and orientation heterogeneities present within a voxel. 

Rui Zeng¹, Jinglei Lv², He Wang³, Luping Zhou², Michael Barnett², Fernando Calamante², and Chenyu Wang^2
1School of Biomedical Engineering, The University of Sydney, Sydney, Australia, ²The University of Sydney, Sydney, Australia, ³Fudan University, Shanghai, China

In this study, a deep learning model called FODSRM was developed for fiber orientation distribution (FOD) super-resolution, which enhances single-shell low-angular-resolution FOD computed from clinic-quality dMRI data (e.g., 32 directions b=1000) to obtain the super-resolved high-angular resolution quality that would have been produced from advanced research scanners (e.g., multi-shell HARDI data). The results demonstrate that the super-resolved FOD data generated by the proposed method can generate high-definition structural connectome from clinical acquisition protocols, even when applied to data from a protocol not included in the trained dataset.

Juan Luis Villarreal Haro^1,2, Gabriel Girard^1,3,4, Jean Philippe Thiran^1,4, and Alonso Ramírez-Manzanares^2
1Signal Processing Lab (LTS5), École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ²Computer Science Department, CIMAT, Centro de Investigación en Matemáticas, Guanajuato, Mexico, ³CIBM Center for BioMedical Imaging, Lausanne, Switzerland, ⁴Radiology Department, Centre Hospitalier Universitaire Vaudois and University of Lausanne, Lausanne, Switzerland

In this work, we study the effect of the reconstruction pipeline s on the reproducibility and sensitivity of the DW-MRI brain connectivity graphs. The brain database we analyze contains several scan repetitions that allow us to characterize the impact of different reconstruction pipeline parameters on the connectomes' topology. We use a novel methodology to detect robust graph communities and show the level of reproducibility of the brain structure/biomarkers. Moreover, we proposed a tool to identify the less-robust graph nodes that suffer the effects of the reconstruction pipelines' variability.

Frederick C. Damen¹, Alessandro Scotti¹, Frederick W. Damen², Nitu Saran¹, Tibor Valyi-Nagy³, Mirko Vukelich¹, and Kejia Cai^1
1Radiology, University of Illinois at Chicago, Chicago, IL, United States, ²Biomedical Engineering, Purdue University, West Lafayette, IN, United States, ³Pathology, University Of Illinois at Chicago, Chicago, IL, United States

In physiological and pathological conditions in which multiple underlying tissue properties are expected to vary the diffusion weighted signal, the diagnostic value of conventional apparent diffusion coefficient (ADC) notably decreases. The proposed method extracts four separate distributions, i.e., modes, of diffusion weighted signal, which includes flow and unimpeded, hindered, and restricted diffusion, thereby increasing the precision of quantification of the hindered diffusion in grey and white matters.

Marshall Axel Dalton¹, Arkiev D'Souza², Jinglei Lv¹, and Fernando Calamante^3
1School of Biomedical Engineering, Faculty of Engineering, University of Sydney, Sydney, Australia, Sydney, Australia, ²DVC Research, Brain and Mind Centre, University of Sydney, Sydney, Australia, Sydney, Australia, ³Sydney Imaging, University of Sydney, Sydney, Australia, Sydney, Australia

The hippocampus is a brain structure central to a broad range of cognitive functions including episodic memory but we know surprisingly little about how different parts of the human hippocampus anatomically connect with cortical regions to support key functions such as memory. We combined high-quality data from the Human Connectome Project with cutting-edge fibre-tracking methods to quantitatively characterise structural connectivity (SC) between the anterior/middle/posterior portions of the hippocampus and whole brain. We also mapped the distribution of endpoints within the hippocampus for streamlines connecting from cortical regions. Our results provide key contributions to ongoing efforts to characterise human hippocampal SC.

Maria Eugenia Caligiuri¹, Andrea Quattrone², Alessandro Mechelli², and Aldo Quattrone^1
1Neuroscience Research Center, University "Magna Graecia", Catanzaro, Italy, ²Institute of Neurology, University "Magna Graecia", Catanzaro, Italy

We propose a novel, semi-automated approach to assess corpus callosum integrity at the individual level, based on the spatial distribution of the principal diffusion direction orientation. We applied the method to the clinical challenge of differentiating normal pressure hydrocephalus (iNPH, N=23) from progressive supranuclear palsy (PSP, N=27) and Alzheimer’s Disease (AD, N=35), whose shared clinical and radiological features can make early differential diagnosis difficult. We found differential involvement of PDD distribution across neurodegenerative (AD, PSP) and non-degenerative diseases (such as iNPH), a finding that could shed light on the different mechanisms that underlie tissue damage.
 

Alberto De Luca^1,2, Suheyla Cetin Karayumak³, Alexander Leemans², Yogesh Rathi³, Stephan Swinnen^4,5, Jolien Gooijers^4,5, Amanda Clauwaert^4,5, Roald Bahr⁶, Stian Bahr Sandmo⁶, Nir Sochen^7,8, David Kaufmann⁹, Marc Muehlmann¹⁰, Geert-Jan Biessels¹, Inga K Koerte^3,11, and Ofer Pasternak^3
1Department of Neurology, University Medical Center Utrecht, Utrecht, Netherlands, ²PROVIDI Lab, Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands, ³Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States, ⁴Movement Control and Neuroplasticity Research Group, KU Leuven, Leuven, Belgium, ⁵Brain Institute, KU Leuven, Leuven, Belgium, ⁶Oslo Sports Trauma Research Center, Norwegian School of Sport Sciences, Oslo, Norway, ⁷Department of Applied Mathematics, Tel Aviv University, Tel Aviv, Israel, ⁸Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel, ⁹Department of Radiology, Charite University Hospital, Berlin, Germany, ¹⁰Department of Radiology, LMU Munich, Munich, Germany, ¹¹cBRAIN, Department of Child and Adolescent Psychiatry, LMU Munich, Munich, Germany

Diffusion Kurtosis Imaging (DKI) metrics computed from diffusion MRI (dMRI) are affected by different acquisition protocols and scanner properties which limits their implementation in multicentric studies. We investigated whether harmonizing multi-shell dMRI with the rotation invariant spherical harmonics (RISH) method allows to remove cross-site differences in DKI metrics while retaining longitudinal effects. 46 subjects underwent a longitudinal two-shell dMRI protocol in 3 imaging sites. Our results show that the RISH method removes inter-site differences both at whole-brain and voxel-level while maintaining the effect-size of longitudinal changes, and is thus promising for the implementation of multi-site DKI studies. 

Ye Wu¹, Sahar Ahmad¹, Lei Ma¹, Erkun Yang¹, and Pew-Thian Yap^1
1Biomedical Research Imaging Center, University of North Carolina, Chapel Hill, Chapel Hill, NC, United States

Subsampling from a set of diffusion gradient directions is useful for evaluation of image acquisition and reconstruction strategies. However, a challenge associated with gradient subsampling is the requirement to ensure uniform distribution of a predetermined number of subsampled directions. Here, we introduce a method for near-uniform subsampling for an arbitrary target number of gradient directions.

Ana R Fouto¹, Rita G Nunes¹, Amparo Ruiz-Tagle¹, Marc Golub¹, Inês Esteves¹, Athanasios Vourvopoulos¹, Raquel Gil-Gouveia², Andreia C Freitas¹, Nuno A Silva³, and Patrícia Figueiredo^1
1ISR-Lisboa/LARSyS and Department of Bioengineering, Instituto Superior Técnico, Universidade de Lisboa, Lisbon, Portugal, ²Neurology Department, Hospital da Luz, Lisbon, Portugal, ³Learning Health, Hospital da Luz, Lisbon, Portugal

Multi-shell diffusion MRI sampling schemes require long acquisition times and are hence potentially compromised by poorer subject compliance. There is therefore a need to consider strategies to enable shortening the exam to ensure clinical feasibility, while maintaining cross-exam comparability. We analyzed the impact of subsampling a multi-shell scheme on the estimation of diffusion maps. In particular, we evaluated its effect on histogram-derived metrics computed over skeletonized maps of DTI and DKI parameters. Several  metrics showed a significant effect of subsampling. This should be carefully considered depending on the effect size we are trying to measure on a patient’s population.

Hu Cheng^1
1Indiana University, Bloomington, IN, United States

We propose a novel but simple method to evaluate noise/signal leaking in PCA-based denoising for diffusion MRI. This method was applied on MP-PCA and LPCA on a human dataset acquired with the ABCD protocol. The results show that MP-PCA is more conservative in removing the noise than LPCA and therefore has little signal leaking. On the other hand, LPCA has less noise leaking as it removed more noise than MP-PCA. Our proposed method can be an effective tool in assessing denoising performance in practice. 

Alberto De Luca^1,2, Andrada Ianus^3,4, Alexander Leemans², Marco Palombo⁵, Hui G Zhang⁵, Daniel C Alexander⁵, Markus Nilsson⁶, Geert-Jan Biessels¹, Mauro Zucchelli⁷, Matteo Frigo⁷, Enes Albay^7,8, Sara Sedlar⁷, Abib Alimi⁷, Samuel Deslauriers-Gauthier⁷, Rachid Deriche⁷, Rutger Fick⁹, Maryam Afzali¹⁰, Tomasz Pieciak^11,12, Fabian Bogusz¹¹, Santiago Aja-Fernandez¹², Evren Özarslan^13,14, Derek K Jones¹⁰, Haoze Chen¹⁵, Mingwu Jin¹⁶, Zhijie Zhang¹⁵, Fengxiang Wang¹⁵, Vishwesh Nath¹⁷, Prasanna Parvathaneni¹⁸, Jan Morez¹⁹, Jan Sijbers¹⁹, Ben Jeurissen¹⁹, Shreyas Fadnavis²⁰, Stefan Endres²¹, Ariel Rokem²², Eleftherios Garyfallidis²⁰, Irina Sanchez²³, Vesna Prchkovska²³, Paulo Rodrigues²³, Bennett A Landman²⁴, and Kurt G Schilling^24
1Department of Neurology, University Medical Center Utrecht, Utrecht, Netherlands, ²PROVIDI Lab, Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands, ³Champalimaud Research, Champalimaud Centre for the Unknown, Lisbon, Portugal, ⁴Centre for Medical Imaging Computing, University College London, London, United Kingdom, ⁵Centre for Medical Image Computing, University College London, London, United Kingdom, ⁶Clinical Science, Department of Radiology, Lund University, Lund, Sweden, ⁷Inria Sophia Antipolis – Méditerranée, Université Côte d'Azur, Sophia Antipolis, France, ⁸Istanbul Technical University, Instanbul, Turkey, ⁹Therapanacea, Paris, France, ¹⁰Cardiff University Brain Research, Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, United Kingdom, ¹¹AGH University of Science and Technology, Krakow, Poland, ¹²LPI, ETSI Telecomunicacion, Universidad de Valladolid, Valladolid, Spain, ¹³Department of Biomedical Engineering, Linköping University, Linköping, Sweden, ¹⁴Center for Medical Image Science and Visualization, Linköping University, Linköping, Sweden, ¹⁵School of Instruments and Electronics, North University of China, Taiyuan, China, ¹⁶Department of Physics, University of Texas at Arlington, Arlington, TX, United States, ¹⁷NVIDIA Corporation, Bethesda, MD, United States, ¹⁸National Institute of Neurological Disorders and Stroke, National Institute of Health, Bethesda, MD, United States, ¹⁹Imec-Vision Lab, Department of Physics, University of Antwerp, Antwerp, Belgium, ²⁰Intelligent Systems Engineering, Indiana University Bloomington, Bloomington, IN, United States, ²¹Leibniz Institute for Materials Engineering - IWT, University of Bremen, Bremen, Germany, ²²Department of Psychology and the eScience Institute, University of Washington, Seattle, WA, United States, ²³QMENTA Inc, Barcelona, Spain, ²⁴Institute of Imaging Science, Vanderbilt University, Nashville, TN, United States

The acquisition and modelling of diffusion MRI (dMRI) data offer many opportunities to explore the organization of the brain. A variety of methods has been proposed to this end, but their generalizability to different diffusion encodings and stronger gradients remains unknown. In the MEMENTO challenge, we asked participants to predict dMRI signals collected with single, double and double oscillating diffusion encodings in combination with a variety of weighting parameters. We received eighty-three submissions ranging from signal representations to multicompartment models and deep learning-based predictions, which offer a unique perspective to investigate the status quo of the field.

David H Gultekin¹, Peter H Siegel^2,3, John T Vaughan¹, and John C Gore^4
1Zuckerman Mind Brain Behavior Institute, Columbia University, New York, NY, United States, ²Jet Propulsion Laboratory, National Aeronautics and Space Administration, Pasadena, CA, United States, ³Department of Electrical Engineering, California Institute of Technology, Pasadena, CA, United States, ⁴Vanderbilt University Institute of Imaging Science, Nashville, TN, United States

A method is introduced to measure the absorption coefficient and the thermal diffusion coefficient of ex vivo brain tissue exposed to radiofrequency radiation emitted from a half wavelength dipole antenna and a rectangular waveguide using MRI and thermocouples. By analyzing the spatial and temporal variations of thermal gradients, it is possible to measure simultaneously both the absorption coefficient and the thermal diffusion coefficient in brain tissue in a single experiment.