David Romascano1,2, Muhamed Barakovic1, Jonathan Rafael-Patino1, Tim Bjørn Dyrby2,3, Jean-Philippe Thiran1,4, Alessandro Daducci1,4,5. 1. Signal Processing Laboratory (LTS5), École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Vaud, Switzerland. 2. Danish Research Centre for Magnetic Resonance, Center for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark. 3. Department of Applied Mathematics and Computer Science, Technical University of Denmark, Kongens Lyngby, Denmark. 4. Department of Radiology, University Hospital Center (CHUV) and University of Lausanne (UNIL), Lausanne, Vaud, Switzerland. 5. Computer Science Department, University of Verona, Verona, Italy.
Abstract
PURPOSE: Non-invasive axon diameter distribution (ADD) mapping using diffusion MRI is an ill-posed problem. Current ADD mapping methods require knowledge of axon orientation before performing the acquisition. Instead, ActiveAx uses a 3D sampling scheme to estimate the orientation from the signal, providing orientationally invariant estimates. The mean diameter is estimated instead of the distribution for the solution to be tractable. Here, we propose an extension (ActiveAxADD ) that provides non-parametric and orientationally invariant estimates of the whole distribution. THEORY: The accelerated microstructure imaging with convex optimization (AMICO) framework accelerates mean diameter estimation using a linear formulation combined with Tikhonov regularization to stabilize the solution. Here, we implement a new formulation (ActiveAxADD ) that uses Laplacian regularization to provide robust estimates of the whole ADD. METHODS: The performance of ActiveAxADD was evaluated using Monte Carlo simulations on synthetic white matter samples mimicking axon distributions reported in histological studies. RESULTS: ActiveAxADD provided robust ADD reconstructions when considering the isolated intra-axonal signal. However, our formulation inherited some common microstructure imaging limitations. When accounting for the extra axonal compartment, estimated ADDs showed spurious peaks and increased variability because of the difficulty of disentangling intra and extra axonal contributions. CONCLUSION: Laplacian regularization solves the ill-posedness regarding the intra axonal compartment. ActiveAxADD can potentially provide non-parametric and orientationally invariant ADDs from isolated intra-axonal signals. However, further work is required before ActiveAxADD can be applied to real data containing extra-axonal contributions, as disentangling the 2 compartment appears to be an overlooked challenge that affects microstructure imaging methods in general.
PURPOSE: Non-invasive axon diameter distribution (ADD) mapping using diffusion MRI is an ill-posed problem. Current ADD mapping methods require knowledge of axon orientation before performing the acquisition. Instead, ActiveAx uses a 3D sampling scheme to estimate the orientation from the signal, providing orientationally invariant estimates. The mean diameter is estimated instead of the distribution for the solution to be tractable. Here, we propose an extension (ActiveAxADD ) that provides non-parametric and orientationally invariant estimates of the whole distribution. THEORY: The accelerated microstructure imaging with convex optimization (AMICO) framework accelerates mean diameter estimation using a linear formulation combined with Tikhonov regularization to stabilize the solution. Here, we implement a new formulation (ActiveAxADD ) that uses Laplacian regularization to provide robust estimates of the whole ADD. METHODS: The performance of ActiveAxADD was evaluated using Monte Carlo simulations on synthetic white matter samples mimicking axon distributions reported in histological studies. RESULTS: ActiveAxADD provided robust ADD reconstructions when considering the isolated intra-axonal signal. However, our formulation inherited some common microstructure imaging limitations. When accounting for the extra axonal compartment, estimated ADDs showed spurious peaks and increased variability because of the difficulty of disentangling intra and extra axonal contributions. CONCLUSION: Laplacian regularization solves the ill-posedness regarding the intra axonal compartment. ActiveAxADD can potentially provide non-parametric and orientationally invariant ADDs from isolated intra-axonal signals. However, further work is required before ActiveAxADD can be applied to real data containing extra-axonal contributions, as disentangling the 2 compartment appears to be an overlooked challenge that affects microstructure imaging methods in general.
Authors: Mariam Andersson; Hans Martin Kjer; Jonathan Rafael-Patino; Alexandra Pacureanu; Bente Pakkenberg; Jean-Philippe Thiran; Maurice Ptito; Martin Bech; Anders Bjorholm Dahl; Vedrana Andersen Dahl; Tim B Dyrby Journal: Proc Natl Acad Sci U S A Date: 2020-12-21 Impact factor: 12.779