pith. sign in

arxiv: 2605.23309 · v1 · pith:MZBLHHGLnew · submitted 2026-05-22 · ⚛️ physics.flu-dyn

Full-component reconstruction of three-dimensional fluid stress tensors

Shunsuke Kumagai , Shun Miyatake , Ryusuke Cho , William Kai Alexander Worby , Masanori Naito , Takahiro Ushioku , Masanobu Horie , Yoshiyuki Tagawa This is my paper

Pith reviewed 2026-05-25 03:17 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn
keywords fluid stress tensorphotoelastic tomography3D reconstructionphysics-informedtensor tomographyflow birefringenceNavier-Stokesunsupervised method
0
0 comments X

The pith

U-FlowPET reconstructs all six components of the 3D fluid stress tensor from optical projections by enforcing momentum balance and continuity.

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

Optical measurements of fluid stress yield only path-integrated data, leaving an underdetermined problem in which two observables must recover six independent tensor components. The paper introduces U-FlowPET, an unsupervised framework that combines photoelastic tomography with the incompressible Navier-Stokes equations to select stress fields consistent with both the measurements and the governing physics. The method requires no constitutive models, no symmetry assumptions, and no labeled training data. Validation on analytical pipe flow recovers every component with normalized mean absolute error below 4 percent; the same procedure succeeds on curved-pipe simulations and noisy experimental recordings. If correct, the approach converts optical flow data into direct, spatially resolved force fields.

Core claim

U-FlowPET is an unsupervised physics-informed framework that integrates photoelastic tomography with the governing equations of fluid mechanics to reconstruct the full 3D stress tensor without relying on constitutive assumptions, geometric symmetry, or labeled training data; it identifies physically admissible stress fields that satisfy momentum balance and continuity while remaining consistent with measured optical projections.

What carries the argument

U-FlowPET, which solves the underdetermined tensor tomography problem by requiring consistency with the incompressible Navier-Stokes equations in addition to the photoelastic effect.

If this is right

  • All six local stress components become experimentally accessible in fully three-dimensional flows without symmetry.
  • Fluid diagnostics can shift from velocity fields to direct quantification of internal forces.
  • Stress-based analysis becomes available for biological flows and soft materials without labeled reference data.
  • The reconstruction remains stable under realistic measurement noise in both simulated and laboratory pipe flows.

Where Pith is reading between the lines

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

  • The same constraint-based approach could be tested on other tensor tomography settings where a known differential equation supplies the missing information.
  • Incorporating the unsteady term of the momentum equation would be a direct next step for time-varying flows.
  • Cross-validation against simultaneous particle-image-velocimetry measurements could check whether the recovered stresses produce the observed accelerations.

Load-bearing premise

The measured optical projections must be produced solely by the photoelastic effect of the stress tensor, and the flow must obey the incompressible Navier-Stokes equations without extra body forces or other unmodeled effects.

What would settle it

Reconstruction of the known analytical stress field in axisymmetric pipe flow that produces normalized mean absolute errors substantially larger than 4 percent for any of the six components.

Figures

Figures reproduced from arXiv: 2605.23309 by Masanobu Horie, Masanori Naito, Ryusuke Cho, Shun Miyatake, Shunsuke Kumagai, Takahiro Ushioku, William Kai Alexander Worby, Yoshiyuki Tagawa.

Figure 1
Figure 1. Figure 1: Concept of U-FlowPET for full-component reconstruction of three-dimensional fluid stress tensors. Auxiliary solid and dashed black lines in the fluid stress tensor fields are used to highlight boundaries and cross-sectional contours. Multi-angle photoelastic measurements provide two optical observables: retardation ∆ and principal orientation ϕ, as path-integrated projections of stress in a flowing birefri… view at source ↗
Figure 2
Figure 2. Figure 2: Full-component reconstruction of three-dimensional fluid stresses using U-FlowPET. Auxiliary solid and dashed black lines highlight boundaries and cross￾sectional contours. For visualization, the stress fields are normalized using the minimum and maximum of the corresponding ground-truth fields. (A) Comparison between ground-truth stress fields and the fields reconstructed by U-FlowPET. Representative comp… view at source ↗
Figure 3
Figure 3. Figure 3: (A) Schematic of the optical measurement principle. Unpolarized light from the source is converted to circularly polarized light using a linear polarizer and a quarter-wave plate before passing through the flowing birefringent fluid. As the light propagates through the stressed medium, the polarization changes due to stress-induced optical anisotropy. The emergent light is detected by a polarization camera… view at source ↗
Figure 4
Figure 4. Figure 4: Architecture of U-FlowPET for full-component reconstruction of three-dimensional fluid stress tensors. Multi-angle photoelastic measurements provide two optical observables—retardation ∆ and principal orientation ϕ—as path-integrated projections of stress. These two inputs are used to reconstruct the full three-dimensional Cauchy stress tensor field (a 3 × 3 tensor with six independent components), represe… view at source ↗
read the original abstract

Forces govern how fluids deform biological tissues, regulate cardiovascular function, and determine the performance and failure of soft materials. Recent advances in flow birefringence, including the use of suspended anisotropic nanomaterials to optically encode stress in fluids, have made direct stress measurement experimentally accessible in projection. However, direct experimental access to all six components of the three-dimensional (3D) fluid stress tensor has remained unattainable because optical measurements provide only path-integrated observables. Recovering local 3D stresses from such data constitutes an intrinsically underdetermined tensor tomography problem, where two optical observables must determine six independent stress components. Here we introduce U-FlowPET, an unsupervised physics-informed framework that integrates photoelastic tomography with the governing equations of fluid mechanics to reconstruct the full 3D stress tensor without relying on constitutive assumptions, geometric symmetry, or labeled training data. Rather than learning from labeled reference stress fields, the method identifies physically admissible stress fields that satisfy momentum balance and continuity while remaining consistent with measured optical projections. We validate the approach using analytical, numerical, and experimental datasets. In axisymmetric pipe flow with an analytical solution, all six stress components are reconstructed with normalized mean absolute errors below 4%. Robust reconstruction is further demonstrated in curved-pipe flow without symmetry assumptions and in experimental pipe-flow data despite measurement noise. By enabling direct 3D stress-field reconstruction from optical data alone, U-FlowPET extends fluid analysis from observing motion to quantifying force and establishes a new framework for stress-based diagnostics in biological flows and functional materials.

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 / 1 minor

Summary. The paper claims to present U-FlowPET, an unsupervised physics-informed framework that reconstructs all six components of the 3D fluid stress tensor from path-integrated optical projections by enforcing momentum balance and continuity from the governing fluid equations, without constitutive models, symmetry assumptions, or labeled data. Validation on an analytical axisymmetric pipe flow yields normalized mean absolute errors below 4% for all components, with additional demonstrations of robustness in curved-pipe flows and noisy experimental data.

Significance. If the results hold, the work would be significant for fluid mechanics by enabling direct 3D stress quantification from optical data alone, extending analysis from kinematics to forces in biological and soft-material flows. The unsupervised enforcement of independent governing equations to resolve the underdetermined tensor tomography problem, without free parameters or invented entities, is a notable strength that supports broad applicability.

major comments (2)
  1. [Methods] Methods: The exact implementation of how momentum balance and continuity residuals are enforced in the unsupervised optimization (including weighting with projection consistency) is not specified, which is load-bearing for confirming that the physics constraints suffice to recover all six independent stress components from two observables.
  2. [Results] Results (analytical validation): The reported NMAE below 4% for all six components is central to the claim, but without the explicit definition of the normalization, the optimization hyperparameters, or sensitivity analysis to the physics residual weights, independent assessment of whether the errors reflect true reconstruction accuracy is not possible.
minor comments (1)
  1. [Abstract] Abstract: The acronym U-FlowPET is introduced without expansion, which may confuse readers unfamiliar with the framework.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects for reproducibility. We address each major comment below and will revise the manuscript accordingly to provide the requested details.

read point-by-point responses

  1. Referee: [Methods] Methods: The exact implementation of how momentum balance and continuity residuals are enforced in the unsupervised optimization (including weighting with projection consistency) is not specified, which is load-bearing for confirming that the physics constraints suffice to recover all six independent stress components from two observables.

    Authors: We agree that the precise implementation details are critical for independent verification. In the revised manuscript, we will expand the Methods section to include the explicit mathematical formulation of the momentum balance and continuity residual terms, the projection consistency loss, and the weighting coefficients in the total unsupervised loss function, along with pseudocode for the optimization procedure. revision: yes

  2. Referee: [Results] Results (analytical validation): The reported NMAE below 4% for all six components is central to the claim, but without the explicit definition of the normalization, the optimization hyperparameters, or sensitivity analysis to the physics residual weights, independent assessment of whether the errors reflect true reconstruction accuracy is not possible.

    Authors: We acknowledge that these details were insufficiently specified. The revised manuscript will define NMAE explicitly (as the mean absolute error normalized by the range of each stress component in the analytical solution), list all optimization hyperparameters (including learning rates, batch sizes, and epoch counts), and add a sensitivity analysis varying the relative weights of the physics residuals to confirm that the reported errors are robust. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation uses independent external constraints

full rationale

The paper's core reconstruction solves an underdetermined inverse problem by minimizing a joint loss that enforces consistency with measured optical projections plus residuals of the incompressible Navier-Stokes equations (momentum balance and continuity). These governing equations are standard, externally derived fluid-mechanics laws that do not depend on the optical data, the reconstruction algorithm, or any fitted parameters from the present work. No step renames a fitted quantity as a prediction, defines a quantity in terms of its own output, or relies on a self-citation chain for a uniqueness claim. Validation against an independent analytical pipe-flow solution further confirms the constraints are not tautological. The method is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The method relies on standard fluid mechanics equations and the photoelastic tomography setup; no new entities introduced. Since only abstract available, details on any optimization parameters or implementation specifics are unknown.

axioms (2)
  • domain assumption Fluid flows satisfy momentum balance and continuity equations
    Used to constrain the reconstructed stress fields
  • domain assumption Optical projections encode stress via photoelastic effect in a known way
    Basis for consistency with measurements

pith-pipeline@v0.9.0 · 5837 in / 1309 out tokens · 37625 ms · 2026-05-25T03:17:43.062271+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

66 extracted references · 66 canonical work pages · 1 internal anchor

  1. [1]

    RJ Adrian, Particle-imaging techniques for experimental fluid mechanics. Annu. Rev. Fluid Mech . 23, 261–304 (1991)

    work page 1991

  2. [2]

    RJ Adrian, Twenty years of particle image velocimetry. Exp. Fluids 39, 159–169 (2005)

    work page 2005

  3. [3]

    F Scarano, Tomographic piv: principles and practice. Meas. Sci. Technol . 24, 012001 (2012)

    work page 2012

  4. [4]

    S Scharnowski, CJ Kähler, Particle image velocimetry - classical operating rules from todayns perspective. Opt. Lasers Eng . 135, 106185 (2020)

    work page 2020

  5. [5]

    A Schroder, D Schanz, 3d lagrangian particle tracking in fluid mechanics. Annu. Rev. Fluid Mech . 55, 511–540 (2023)

    work page 2023

  6. [6]

    PF Davies, Flow-mediated endothelial mechanotransduction. Physiol. Rev . 75, 519–560 (1995) PMID: 7624393

    work page 1995

  7. [7]

    P Davies, Hemodynamic shear stress and the endothelium in cardiovascular pathophysiology. Nat. clinical practice. Cardiovasc. medicine 6, 16–26 (2008)

    work page 2008

  8. [8]

    CA Figueroa, S Baek, CA Taylor, JD Humphrey, A computational framework for fluid-solid-growth modeling in cardiovascular simulations. Comput. Methods Appl. Mech. Eng . 198, 3583–3602 (2009)

    work page 2009

  9. [9]

    AL Marsden, et al., Evaluation of a novel y-shaped extracardiac fontan baffle using computational fluid dynamics. The J. Thorac. Cardiovasc. Surg . 137, 394–403 (2009)

    work page 2009

  10. [10]

    NM Wilson, FR Arko, CA Taylor, Predicting changes in blood flow in patient-specific operative plans for treating aortoiliac occlusive disease. Comput. Aided Surg . 10, 257–277 (2005)

    work page 2005

  11. [11]

    CA Taylor, CA Figueroa, Patient-specific modeling of cardiovascular mechanics. Annu. Rev. Biomed. Eng . 11, 109–134 (2009)

    work page 2009

  12. [12]

    CA Taylor, et al., Predictive medicine: computational techniques in therapeutic decision-making. Comput. Aided Surg . 4, 231–247 (1999)

    work page 1999

  13. [13]

    A Blaeser, et al., Controlling shear stress in 3d bioprinting is a key factor to balance printing resolution and stem cell integrity. Adv. Healthc. Mater . 5, 326–333 (2016)

    work page 2016

  14. [14]

    HQ Xu, JC Liu, ZY Zhang, CX Xu, A review on cell damage, viability, and functionality during 3d bioprinting. Mil. Med. Res . 9, 70 (2022)

    work page 2022

  15. [15]

    Soft Robotics 1, 75–87 (2014)

    AD Marchese, CD Onal, D Rus, Autonomous soft robotic fish capable of escape maneuvers using fluidic elastomer actuators. Soft Robotics 1, 75–87 (2014)

    work page 2014

  16. [16]

    (IEEE), pp

    AD Marchese, R Tedrake, D Rus, Dynamics and trajectory optimization for a soft spatial fluidic elastomer manipulator in 2015 IEEE International Conference on Robotics and Automation (ICRA) . (IEEE), pp. 2528–2535 (2015)

    work page 2015

  17. [17]

    Nature 521, 467–475 (2015)

    D Rus, MT Tolley, Design, fabrication and control of soft robots. Nature 521, 467–475 (2015)

    work page 2015

  18. [18]

    Catheter

    W Liu, et al., Pressure guidewire indicated for coronary fractional flow reserve assessment: Structural design, deficiencies and potential solutions. Catheter. Cardiovasc. Interv . (2025)

    work page 2025

  19. [19]

    Sensors 25, 7178 (2025)

    V Bučinskas, JJ Petronienė, G Vaičiūnas, N Šešok, A Dzedzickis, Integrated polymeric sensors in heart and blood vessel monitoring: A review. Sensors 25, 7178 (2025)

    work page 2025

  20. [20]

    WJ McAfee, H Pih, Scattered-light flow-optic relations adaptable to three-dimensional flow birefringence. Exp. Mech . 14, 385–391 (1974)

    work page 1974

  21. [21]

    Inverse Probl

    H Aben, A Puro, Photoelastic tomography for three-dimensional flow birefringence studies. Inverse Probl . 13, 215–221 (1997). 9

    work page 1997

  22. [22]

    H Aben, A Errapart, L Ainola, J Anton, Photoelastic tomography for residual stress measurement in glass. Opt. Eng . 44, 093601 (2005)

    work page 2005

  23. [23]

    A Errapart, On the technology of photoelastic tomography. Exp. Tech. 31, 19–21 (2007)

    work page 2007

  24. [24]

    HT Jessop, The determination of the separate stresses in three-dimensional stress investigations by the frozen stress method. J. Sci. Instruments 26, 27 (1949)

    work page 1949

  25. [25]

    JF Doyle, On a nonlinearity in flow birefringence. Exp. Mech . 22, 37–38 (1982)

    work page 1982

  26. [26]

    BE Schmidt, et al., Twenty-five years of background-oriented schlieren: Advances and novel applications in AIAA SCITECH 2025 Forum . p. 1567 (2025)

    work page 2025

  27. [27]

    AM Cormack, Representation of a function by its line integrals, with some radiological applications. J. Appl. Phys . 34, 2722–2727 (1963)

    work page 1963

  28. [28]

    R Gordon, R Bender, GT Herman, Algebraic reconstruction techniques (art) for three-dimensional electron microscopy and x-ray photography. J. Theor. Biol . 29, 471–481 (1970)

    work page 1970

  29. [29]

    GN Ramachandran, A V Lakshminarayanan, Three-dimensional reconstruction from radiographs and electron micrographs: application of convolutions instead of fourier transforms. Proc. Natl. Acad. Sci . 68, 2236–2240 (1971)

    work page 1971

  30. [30]

    H Aben, Integrated photoelasticity of axisymmetric bodies. Opt. Eng . 21, 689–695 (1982)

    work page 1982

  31. [31]

    Photoelasticity (Springer-Verlag Tokyo) pp

    HK Aben, Integrated photoelasticity as tensor field tomography. Photoelasticity (Springer-Verlag Tokyo) pp. 243–250 (1986)

    work page 1986

  32. [32]

    H Aben, A Puro, Integrated photoelasticity for axisymmetric flow birefringence studies. Proc. Estonian Acad. Sci. Phys. Math. 42, 7–13 (1993)

    work page 1993

  33. [33]

    JF Doyle, HT Danyluk, Integrated photoelasticity for axisymmetric problems. Exp. Mech. 18, 215–220 (1978)

    work page 1978

  34. [34]

    Acta Mech

    Y Chen, Y Jia, W Shu, Utilizing phase retardation integration method of flow birefringence to analyze the flow with symmetric plane. Acta Mech. Sinica 6, 358–365 (1990)

    work page 1990

  35. [35]

    C Lane, F Baumann, D Rode, T Rösgen, Two-dimensional strain rate imaging study using a polarization camera and birefringent aqueous cellulose nanocrystal suspensions. Exp. Fluids 65, 8 (2024)

    work page 2024

  36. [36]

    W Li, R Juanes, Dynamic imaging of force chains in 3d granular media. Proc. Natl. Acad. Sci . 121, e2318751121 (2024)

    work page 2024

  37. [37]

    Reconstruction of spatially inhomogeneous dielectric tensors via optical tomography

    H Hammer, WRB Lionheart, Reconstruction of spatially inhomogeneous dielectric tensors through optical tomography. arXiv preprint arXiv:physics/0406053 (2004)

    work page internal anchor Pith review Pith/arXiv arXiv 2004

  38. [38]

    Inverse Probl

    WRB Lionheart, V Sharafutdinov, Reconstruction algorithm for the polarization tomography problem with incomplete data. Inverse Probl . 24, 1–18 (2008) MIMS EPrint: 2008.75

    work page 2008

  39. [39]

    Limited Data Problems in X-ray and Polarized Light Tomography,

    D Szotten, “Limited Data Problems in X-ray and Polarized Light Tomography,” PhD thesis, University of Manchester (2011)

    work page 2011

  40. [40]

    Experimental Assessment and Implementation of Photoelastic Tomography,

    SG Abrego Hernández, “Experimental Assessment and Implementation of Photoelastic Tomography,” PhD thesis, University of Sheffield (2019)

    work page 2019

  41. [41]

    L Ainola, H Aben, On the optical theory of photoelastic tomography. J. Opt. Soc. Am. A 21, 1093–1101 (2004)

    work page 2004

  42. [42]

    H Aben, A Errapart, Photoelastic tomography with linear and non-linear algorithms. Exp. Mech . 52, 1179–1193 (2012)

    work page 2012

  43. [43]

    M Raissi, P Perdikaris, G Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys . 378, 686–707 (2019)

    work page 2019

  44. [44]

    Science 367, 1026–1030 (2020)

    M Raissi, A Yazdani, GE Karniadakis, Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations. Science 367, 1026–1030 (2020)

    work page 2020

  45. [45]

    S Cai, et al., Flow over an espresso cup: inferring 3-d velocity and pressure fields from tomographic background oriented schlieren via physics-informed neural networks. J. Fluid Mech . 915 (2021)

    work page 2021

  46. [46]

    IEEE Transactions on Image Process

    KH Jin, MT McCann, E Froustey, M Unser, Deep convolutional neural network for inverse problems in imaging. IEEE Transactions on Image Process . 26, 4509–4522 (2017)

    work page 2017

  47. [47]

    arXiv preprint arXiv:2504.15952 (2025)

    D Igarashi, et al., Reconstruction of three-dimensional fluid stress field via photoelasticity using physics-informed convolutional encoder-decoder. arXiv preprint arXiv:2504.15952 (2025)

    work page arXiv 2025

  48. [48]

    R Sergazinov, M Kramár, Machine learning approach to force reconstruction in photoelastic materials. Mach. Learn. Sci. Technol . 2, 045030 (2021)

    work page 2021

  49. [49]

    B Tao, et al., Photoelastic stress field recovery using deep convolutional neural network. Front. Bioeng. Biotechnol . 10, 818112 (2022)

    work page 2022

  50. [50]

    Bioengineering 6 (2019)

    H Xie, H Shan, G Wang, Deep encoder-decoder adversarial reconstruction (dear) network for 3d ct from few-view data. Bioengineering 6 (2019)

    work page 2019

  51. [51]

    N Shapira, S Bharthulwar, PB Noël, Convolutional encoder-decoder networks for volumetric computed tomography surviews from single- and dual-view topograms. Res. Sq. (2022)

    work page 2022

  52. [52]

    M Yousif, L Yu, S Hoyas, R Vinuesa, H Lim, A deep-learning approach for reconstructing 3d turbulent flows from 2d observation data. Sci. Reports 13 (2023)

    work page 2023

  53. [53]

    SN Comput

    M Matsuo, K Fukami, T Nakamura, M Morimoto, K Fukagata, Reconstructing three-dimensional bluff body wake from sectional flow fields with convolutional neural networks. SN Comput. Sci . 5 (2024)

    work page 2024

  54. [54]

    Cellulose 31, 7405–7420 (2024)

    K Nakamine, Y Yokoyama, WKA Worby, M Muto, Y Tagawa, Flow birefringence of cellulose nanocrystal suspensions in three-dimensional flow fields: revisiting the stress-optic law. Cellulose 31, 7405–7420 (2024)

    work page 2024

  55. [55]

    M Muto, KU Kobayashi, A Sawai, R Umezawa, S Tamano, Development of solid-liquid birefringence method targeting dynamic stress interactions between vascular phantom and blood analogue. Phys. Fluids 37, 081907 (2025)

    work page 2025

  56. [56]

    T Onuma, Y Otani, A development of two-dimensional birefringence distribution measurement system with a sampling rate of 1.3mhz. Opt. Commun . 315, 69–73 (2014)

    work page 2014

  57. [57]

    RB Bird, WE Stewart, EN Lightfoot, Transport Phenomena. (J. Wiley), (2002)

    work page 2002

  58. [58]

    Cellulose 29, 6093–6107 (2022)

    C Lane, D Rode, T Rösgen, Birefringent properties of aqueous cellulose nanocrystal suspensions. Cellulose 29, 6093–6107 (2022)

    work page 2022

  59. [59]

    JD Riera, R Mark, The optical-rotation effect in photoelastic shell analysis. Exp. Mech. 9, 9–16 (1969)

    work page 1969

  60. [60]

    M Horie, N Mitsume, Physics-embedded neural networks: Graph neural PDE solvers with mixed boundary conditions in Advances in Neural Information Processing Systems. (2022)

    work page 2022

  61. [61]

    M Horie, N Mitsume, Graph neural PDE solvers with conservation and similarity-equivariance in Forty-first International Conference on Machine Learning . (2024). 10 Supporting Information for1 Full-component reconstruction of three-dimensional fluid stress tensors2 Shunsuke Kumagai, Shun Miyatake, Ryusuke Cho, William Kai Alexander Worby, Masanori Naito, T...

    work page 2024

  62. [62]

    The solution preparation followed the procedure reported by Muto et al.( 1), with17 appropriate modifications for the present experimental conditions

    Sample preparation and rheological characterization15 The working fluid used in this study consisted of a suspension of cellulose nanocrystals (CNC, 0.3 wt.%) dispersed in an16 aqueous sodium iodide (NaI; 57.6 wt.%). The solution preparation followed the procedure reported by Muto et al.( 1), with17 appropriate modifications for the present experimental c...

    work page

  63. [63]

    Identification of Stress-Optic Coefficients36 Calibration procedure. The stress-optic coefficients were determined by fitting the theoretical retardation distribution predicted37 from the analytical stress field to the experimentally measured retardation profiles.38 As described in the main text, the stress distribution inside the circular pipe was first co...

    work page

  64. [64]

    S3 shows the geometric configuration of the curved pipe used to generate the synthetic dataset

    Geometry of the Curved Pipe Used to Generate Synthetic Data61 Fig. S3 shows the geometric configuration of the curved pipe used to generate the synthetic dataset. The pipe has a circular62 cross section with an inner diameter of 4 mm and a total length of 100 mm.63 The geometry of the curved section is illustrated from two viewing directions obtained by r...

    work page

  65. [65]

    The detailed network architecture72 is illustrated in Fig

    Detailed Architecture of the U-FlowPET Network71 The proposed U-FlowPET model is based on a 3D convolutional neural network (3D CNN). The detailed network architecture72 is illustrated in Fig. S4.73 The input layer has dimensions 128 × 128 × 4 × 12, representing multi-angle optical observations. For each of the 12 viewing74 directions, the input channels ...

    work page

  66. [66]

    M Muto, KU Kobayashi, A Sawai, R Umezawa, S Tamano, Development of solid ⚶liquid birefringence method targeting98 dynamic stress interactions between vascular phantom and blood analogue. Phys. Fluids 37, 081907 (2025).99 Kumagai et al. 5 of 5

    work page 2025