Kwok-Shing Chan¹, Tae Hyung Kim^2,3, Berkin Bilgic^2,3, and José P Marques^1
1Donders Institute for Brain, Cognition and Behaviour, Nijmegen, Netherlands, ²Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ³Department of Radiology, Harvard Medical School, Boston, MA, United States

Myelin water imaging using multi-compartment relaxometry (MCR-MWI) improves the GRE-MWI robustness and accuracy but suffered from slow processing speed. In this study, we incorporate both supervised and self-supervised machine learning for fast MCR-MWI that is generalisable to a wide range of acquisition parameters without the need to re-train the network. We demonstrate its application on single compartment fitting and MCR-MWI. Results show that the proposed method can produce comparable high SNR results with a 62-fold shorter processing time.

Jessica Schäper^1,2 and Oliver Bieri^1,2
1Department of Biomedical Engineering, University of Basel, Basel, Switzerland, ²Division of Radioligical Physics, Department of Radiology, University of Basel Hospital, Basel, Switzerland

Myelin water fraction (MWF) imaging is an important tool to investigate demyelination in WM. When using GRE sequences, the method suffers from T₂* effects, leading to a non-exponential behaviour. This can potentially influence the calculation of MWF. At low fields susceptibility effects are less and the assumption of biexponential T₂-decay is more accurate. Here, the feasibility of MWF at 0.55 T, using biexponential fitting was demonstrated.

Björn Lampinen^1,2, Filip Szczepankiewicz^3,4, and Markus Nilsson^3
1Clinical Sciences Lund, Medical Radiation Physics, Lund University, Lund, Sweden, ²Skåne University Hospital, Lund, Sweden, ³Clincial Sciences Lund, Diagnostic Radiology, Lund University, Lund, Sweden, ⁴Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States

White matter pathology is characterized by demyelination and axonal loss. Diffusion MRI and relaxometry cannot separate these processes because they are primarily sensitive to axons and myelin, respectively. We used diffusion-T₁-T₂-relaxation MRI with tensor-valued encoding to support a twelve-parameter microstructure model including myelin. Results yielded brain parameter maps and values of the axonal and myelin volume fractions and the g-ratio that were in accordance with previous results from histology. The proposed approach can theoretically disambiguate between demyelination and axonal loss and could thus become a valuable tool for assessing disease severity and treatment response in pathologies that affect white matter.

Giorgia Milotta¹, Nadège Corbin^1,2, Christian Lambert¹, Antoine Lutti³, Siawoosh Mohammadi^4,5, and Martina Callaghan^1
1Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, ²Centre de Résonance Magnétique des Systèmes Biologiques, UMR5536, CNRS/University Bordeaux, Bordeaux, France, ³Laboratory for Research in Neuroimaging, Department for Clinical Neuroscience, , Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ⁴Department of Systems Neurosciences, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, ⁵Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany

The apparent transverse relaxation rate ($$$R_{2}^{*}$$$) is influenced by biological features, e.g. iron and myelin content. However confounding factors, such as flip angle (α) and fibre orientation dependence, hinder interpretation. Multi-α data acquired as part of a comprehensive multi-parameter mapping approach can be used to mitigate these confounds. Here we explored how best to do so in vivo at 7T while additionally considering reproducibility. The ESTATICS approach, which assumes a common decay across flip angles, reduced these dependencies. $$$\hat{R_{2}^{*}}$$$, the α-independent component of a heuristic linear model, reduced both dependencies further but was less reproducible than ESTATICS.

Lin Chen^1,2, Deng Mao², Guillaume Gilbert³, Maarten Versluis⁴, Peter van Zijl^1,2, and Xu Li^1,2
1Department of Radiology and Radiological Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, ³MR Clinical Science, Philips Canada, Mississauga, ON, Canada, ⁴MRI Clinical Scientist Neuroscience, Philips Healthcare, Best, Netherlands

We investigated the dependence of local frequency difference Δf and R₂^* on the fiber orientation with respect to the static field in selected parallel-fiber and crossing-fiber regions at 3T and 7T. Δf and R₂^* showed consistent relationship (sin²θ and sin⁴θ ) with average ROI-based fiber-to-field angle θ at 3T and 7T. Larger amplitude of variations in Δf and R₂^* with respect to  θ were observed in parallel-fiber compared with crossing-fiber regions indicating non-negligible contributions from fiber orientation dispersion to the nonlinear gradient echo signal evolution and related MRI measures sensitive to white matter microstructure.

Jose Pedro Manzano Patron^1,2, Theodore Kypraios³, and Stamatios N Sotiropoulos^4,5
1Sir Peter Mansfield Imaging Centre, Mental Health and Clinical Neurosciences, School of Medicine, University of Nottingham, Nottingham, United Kingdom, Nottingham, United Kingdom, ²Precision Imaging Beacon, University of Nottingham, Nottingham, United Kingdom, ³School of Mathematics, University of Nottingham, Nottingham, United Kingdom, ⁴Sir Peter Mansfield Imaging Centre, Mental Health and Clinical Neurosciences, School of Medicine, University of Nottingham, Nottingham, United Kingdom, ⁵Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom

Inference in imaging-based biophysical modelling provides a principled way of estimating model parameters, but also assessing confidence/uncertainty on results, quantifying noise effects and aiding experimental design. Traditional approaches in neuroimaging can either be very computationally expensive (e.g., Bayesian) or suitable to only certain assumptions (e.g., bootstrapping). We present a simulation-based inference approach to estimate diffusion MRI model parameters and their uncertainty. This novel framework trains a neural network to learn a Bayesian model inversion, allowing inference given unseen data. Results show a high level of agreement with conventional Markov-Chain-Monte-Carlo estimates, while offering 2-3 orders of magnitude speed-ups and inference amortisation.

Sebastian Hübner¹, Lisa Novello¹, Giacomo Tomezzoli¹, and Jorge Jovicich^1
1CIMeC, Center for Mind/Brain Sciences, University of Trento, Rovereto (Trento), Italy

The thalamus is a primary station for information processing within the brain. It is composed of a collection of nuclei with different histological and functional properties, classified as First Order (FO, processing sensory input) or Higher Order (HO, playing associative roles). We used diffusion-weighted MRI (dMRI) to investigate microstructural properties of FO and HO thalamic nuclei, focusing on fractional anisotropy, mean diffusivity, and mean kurtosis derived from Diffusion Kurtosis Imaging (DKI). This initial evidence suggests that DKI might capture distinctive features of FO and HO thalamic nuclei, and paves the way to further studies investigating nucleus-specific microstructural effects with DKI.