pith. sign in

arxiv: 1907.06139 · v1 · pith:LS5ALCOInew · submitted 2019-07-13 · ⚛️ physics.med-ph · physics.bio-ph

Optimal experimental design for biophysical modelling in multidimensional diffusion MRI

Santiago Coelho , Jose M. Pozo , Sune N. Jespersen , Alejandro F. Frangi This is my paper

Pith reviewed 2026-05-24 21:52 UTC · model grok-4.3

classification ⚛️ physics.med-ph physics.bio-ph
keywords optimal experimental designmultidimensional diffusion MRIStandard ModelFisher information matrixbiophysical modellingdiffusion tensor encodingparameter estimationacquisition optimization
0
0 comments X

The pith

Maximizing the determinant of the Fisher information matrix averaged over the full Standard Model parameter space identifies objective optimal designs for multidimensional diffusion MRI sequences.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper creates a framework to select the best sets of diffusion measurements for recovering the parameters of the Standard Model of brain tissue microstructure. Traditional pulsed-gradient pairs leave the inverse problem ill-posed under clinical noise levels, but multidimensional encoding that mixes linear and planar b-tensors supplies enough information when enough measurements are taken. The authors replace the usual dependence on fixed tissue values by averaging the determinant of the Fisher information matrix across the entire possible range of Standard Model parameters. The resulting scalar depends only on the acquisition configuration, can be maximized under hardware or other practical constraints, and is computed explicitly for the family of b-tensors that share fixed eigenvectors.

Core claim

We develop a framework for optimal experimental design of multidimensional dMRI sequences applicable to the SM. This framework is based on maximising the determinant of the Fisher information matrix, which is averaged over the full SM parameter space. This averaging provides a fairly objective information metric tailored for the expected signal but that only depends on the acquisition configuration. The optimisation of this metric can be further restricted to any subclass of desirable design constraints like, for instance, hardware-specific constraints. In this work, we compute the optimal acquisitions over the set of all b-tensors with fixed eigenvectors.

What carries the argument

The determinant of the Fisher information matrix averaged over the full Standard Model parameter space, which produces an objective acquisition-quality metric that depends only on the chosen b-tensor set.

If this is right

  • Optimal b-tensor sets can be identified without first choosing specific tissue parameter values.
  • The same averaged metric can be maximized inside any chosen subclass of acquisitions, including hardware-limited ones.
  • Combining linear and planar tensor encodings yields enough information to estimate all SM parameters once the design is optimized.
  • The resulting protocols are expected to improve parameter recovery under the noise levels typical of clinical scans.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The averaging technique could be applied to other biophysical models that also suffer from parameter-dependent information content.
  • Scan-time savings may become possible if the optimized designs reach target precision with fewer measurements than current protocols.
  • Similar objective metrics might improve experimental design in other noisy inverse problems where ground-truth parameter distributions are broad.

Load-bearing premise

Averaging the Fisher information matrix determinant over the entire SM parameter space yields an acquisition-dependent metric that is appropriate and objective for clinical protocols with noise, without needing to specify particular tissue parameter values in advance.

What would settle it

A Monte-Carlo simulation or phantom experiment in which acquisitions chosen by the averaged metric produce lower variance or bias in recovered SM parameters than non-optimized multidimensional or conventional schemes, across a wide range of ground-truth parameter values and realistic noise levels.

Figures

Figures reproduced from arXiv: 1907.06139 by Alejandro F. Frangi, Jose M. Pozo, Santiago Coelho, Sune N. Jespersen.

Figure 1
Figure 1. Figure 1: Superquadric tensor glyphs arranged in a barycentric ternary diagram [9] ac￾cording to their linear, planar, and spherical components (LTE, PTE, and STE). Two degrees of freedom define the tensor shape, and an extra one is needed for its size. Multidimensional dMRI was proposed in [8], assuming an underlying diffu￾sion tensor distribution, to disentangle orientation dispersion and microstruc￾tural anisotro… view at source ↗
Figure 2
Figure 2. Figure 2: Optimal experimental designs considering each set of constraints (C1 left, C2 right). Individual dots represent the shapes of each tensor in each B-set (along XY plane in a triangular grid based on [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
read the original abstract

Computational models of biophysical tissue properties have been widely used in diffusion MRI (dMRI) research to elucidate the link between microstructural properties and MR signal formation. For brain tissue, the research community has developed the so-called Standard Model (SM) that has been widely used. However, in clinically applicable acquisition protocols, the inverse problem that recovers the SM parameters from a set of MR diffusion measurements using pairs of short pulsed field gradients was shown to be ill-posed. Multidimensional dMRI was shown to solve this problem by combining linear and planar tensor encoding data. Given sufficient measurements, multiple choices of b-tensor sets provide enough information to estimate all SM parameters. However, in the presence of noise, some sets will provide better results. In this work, we develop a framework for optimal experimental design of multidimensional dMRI sequences applicable to the SM. This framework is based on maximising the determinant of the Fisher information matrix, which is averaged over the full SM parameter space. This averaging provides a fairly objective information metric tailored for the expected signal but that only depends on the acquisition configuration. The optimisation of this metric can be further restricted to any subclass of desirable design constraints like, for instance, hardware-specific constraints. In this work, we compute the optimal acquisitions over the set of all b-tensors with fixed eigenvectors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript develops a framework for optimal experimental design of multidimensional dMRI sequences applicable to the Standard Model (SM) of brain tissue. It maximizes the determinant of the Fisher information matrix averaged over the full SM parameter space, yielding an acquisition-dependent information metric claimed to be objective and independent of specific tissue parameter values. The optimization is performed over b-tensors with fixed eigenvectors, subject to hardware constraints, to address the ill-posed inverse problem in standard pulsed-gradient acquisitions.

Significance. If the central claims hold after addressing the noted issues, the work would provide a principled, prior-averaged D-optimal design method for multidimensional diffusion encodings. This could improve robustness of SM parameter estimation in noisy clinical protocols by identifying b-tensor sets that perform well across the expected range of tissue properties without requiring case-specific tuning. The approach builds on established Fisher-information techniques but applies them to the SM in a way that directly tackles the dimensionality and noise sensitivity of multidimensional dMRI.

major comments (2)
  1. [Abstract] Abstract: The claim that averaging det(I(θ)) over the full SM parameter space produces a metric that 'only depends on the acquisition configuration' and is 'fairly objective' rests on an implicit uniform prior over the chosen parameter bounds (fractions, diffusivities, dispersion). The manuscript provides no justification for these bounds from histological or population data, nor any sensitivity analysis showing that the resulting optima are stable under reasonable changes to the integration limits. This is load-bearing for the 'objective' framing.
  2. [Abstract] Abstract and methods description: The framework is said to compute optimal acquisitions, yet the provided text contains no reported validation (e.g., Monte-Carlo simulations of parameter recovery error, comparison against standard linear-tensor or random b-tensor designs, or Cramér-Rao bound checks on the optimized versus baseline protocols). Without such evidence the practical advantage of the averaged criterion remains unquantified.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address each major comment below and will revise the manuscript to incorporate clarifications and additional analyses where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that averaging det(I(θ)) over the full SM parameter space produces a metric that 'only depends on the acquisition configuration' and is 'fairly objective' rests on an implicit uniform prior over the chosen parameter bounds (fractions, diffusivities, dispersion). The manuscript provides no justification for these bounds from histological or population data, nor any sensitivity analysis showing that the resulting optima are stable under reasonable changes to the integration limits. This is load-bearing for the 'objective' framing.

    Authors: We agree that the averaging procedure implicitly adopts a uniform distribution over the integration limits, and that these limits require explicit justification to support the claim of a fairly objective metric. The bounds used in the manuscript were selected from ranges commonly employed in the Standard Model literature for brain tissue, but we acknowledge the absence of direct histological or population-based references and sensitivity testing. We will revise the methods section to cite supporting literature for the parameter ranges and add a sensitivity analysis demonstrating stability of the resulting optimal acquisitions under reasonable perturbations of the bounds. revision: yes

  2. Referee: [Abstract] Abstract and methods description: The framework is said to compute optimal acquisitions, yet the provided text contains no reported validation (e.g., Monte-Carlo simulations of parameter recovery error, comparison against standard linear-tensor or random b-tensor designs, or Cramér-Rao bound checks on the optimized versus baseline protocols). Without such evidence the practical advantage of the averaged criterion remains unquantified.

    Authors: The manuscript's core contribution is the formulation of the prior-averaged D-optimality criterion and its use to identify optimal b-tensor sets under fixed-eigenvector and hardware constraints. While the optimization results are shown, we concur that the text does not include Monte-Carlo validation or direct comparisons of estimation performance. We will add a dedicated validation section with Monte-Carlo simulations of parameter recovery, comparisons against linear-tensor and random b-tensor protocols, and Cramér-Rao bound evaluations to quantify the practical gains. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard D-optimality criterion applied to SM

full rationale

The paper defines its optimality metric directly as the average of det(Fisher information) over the SM parameter space and then optimizes acquisitions under that definition. This is a methodological choice with no reduction of the claimed result to a fitted parameter, self-referential definition, or load-bearing self-citation. The derivation chain is self-contained: the Fisher matrix is computed from the forward signal model, averaged by explicit integration, and maximized; none of these steps presuppose the final optimal design. No equations or claims in the provided text exhibit the enumerated circular patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the Standard Model being an appropriate description and on the validity of the Fisher information matrix as an information measure for design; no free parameters or new entities are introduced in the abstract.

axioms (2)
  • domain assumption The Standard Model parameters can be recovered from multidimensional dMRI data when sufficient measurements are available.
    Stated as background that multidimensional dMRI solves the ill-posedness of the inverse problem.
  • domain assumption Maximizing the determinant of the averaged Fisher information matrix identifies acquisitions that perform better in the presence of noise.
    Core premise of the optimal design framework.

pith-pipeline@v0.9.0 · 5773 in / 1372 out tokens · 23794 ms · 2026-05-24T21:52:37.060732+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

  1. [1]

    Fundamentals of diffusion MRI physics,

    V. G. Kiselev, “Fundamentals of diffusion MRI physics,” NMR in Biomedicine , vol. 30, pp. 1–18, 2017

    work page 2017

  2. [2]

    Can we use diffusion MRI as a bio-marker of neurodegenerative pro- cesses?,

    Y. Assaf, “Can we use diffusion MRI as a bio-marker of neurodegenerative pro- cesses?,” BioEssays, vol. 30, no. 11-12, pp. 1235–1245, 2008

    work page 2008

  3. [3]

    New modeling and experimental framework to characterize hindered and restricted water diffusion in brain white matter,

    Y. Assaf, T. Blumenfeld-Katzir, Y. Yovel, and P. J. Basser, “New modeling and experimental framework to characterize hindered and restricted water diffusion in brain white matter,” Magnetic Resonance in Medicine , vol. 52, pp. 965–978, 2004

    work page 2004

  4. [4]

    Rotationally-invariant mapping of scalar and orientational metrics of neuronal microstructure with diffu- sion MRI,

    D. S. Novikov, J. Veraart, I. O. Jelescu, and E. Fieremans, “Rotationally-invariant mapping of scalar and orientational metrics of neuronal microstructure with diffu- sion MRI,” NeuroImage, vol. 174, pp. 518 – 538, 2018

    work page 2018

  5. [5]

    Degeneracy in model parameter estimation for multi-compartmental diffusion in neuronal tissue,

    I. O. Jelescu, J. Veraart, E. Fieremans, and D. S. Novikov, “Degeneracy in model parameter estimation for multi-compartmental diffusion in neuronal tissue,” NMR in Biomedicine, vol. 29, pp. 33–47, 2016

    work page 2016

  6. [6]

    Resolving degeneracy in diffusion MRI biophysical model parameter estimation using double diffusion encoding,

    S. Coelho, J. M. Pozo, S. N. Jespersen, D. K. Jones, and A. F. Frangi, “Resolving degeneracy in diffusion MRI biophysical model parameter estimation using double diffusion encoding,” Magnetic Resonance in Medicine , vol. 82, pp. 395–410, 2019

    work page 2019

  7. [7]

    A unique analytical solution of the white matter standard model using linear and planar encodings,

    M. Reisert, V. G. Kiselev, and B. Dhital, “A unique analytical solution of the white matter standard model using linear and planar encodings,” Magnetic Resonance in Medicine, vol. 81, pp. 3819–3825, 2019

    work page 2019

  8. [8]

    q-space trajectory imaging for multidimensional diffusion MRI of the human brain,

    C.-F. Westin, H. Knutsson, O. Pasternak, F. Szczepankiewicz, E. ¨Ozarslan, D. van Westen, C. Mattisson, M. Bogren, L. J. O’Donnell, M. Kubicki, D. Topgaard, and M. Nilsson, “q-space trajectory imaging for multidimensional diffusion MRI of the human brain,” NeuroImage, vol. 135, pp. 345–362, 2016. Title Suppressed Due to Excessive Length 9

    work page 2016

  9. [9]

    Multidimensional diffusion MRI,

    D. Topgaard, “Multidimensional diffusion MRI,” Journal of Magnetic Resonance , vol. 275, pp. 98–113, 2017

    work page 2017

  10. [10]

    Is the biexponential diffusion biexponential?,

    V. G. Kiselev and K. A. Il’yasov, “Is the biexponential diffusion biexponential?,” Magnetic Resonance in Medicine , vol. 57, no. 3, pp. 464–469, 2007

    work page 2007

  11. [11]

    L. L. Scharf, Statistical Signal Processing: Detection, Estimation, and Time Series Analysis. Addison-Wesley Publishing Company, 1991

    work page 1991

  12. [12]

    The Rician distribution of noisy MRI data,

    H. Gudbjartsson and S. Patz, “The Rician distribution of noisy MRI data,” Mag- netic Resonance in Medicine , vol. 34, no. 6, pp. 910–914, 1995

    work page 1995

  13. [13]

    Soma — self-organizing migrating algorithm,

    I. Zelinka, “Soma — self-organizing migrating algorithm,” in New Optimization Techniques in Engineering , ch. 7, pp. 167–217, Springer Berlin Heidelberg, 2004

    work page 2004

  14. [14]

    Ackley, A connectionist machine for genetic hillclimbing

    D. Ackley, A connectionist machine for genetic hillclimbing . Kluwer Academic Publishers, 1987

    work page 1987