Pith. sign in

REVIEW 4 major objections 6 minor 43 references

Pose-Free 3D Quantitative Phase Imaging of Flowing Cellular Populations

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read OmniFHT reconstructs 3D refractive-index maps of flowing cells without knowing their pose, using as few as ten views or 120 degrees of angular coverage.

desk verdict Solid pose-free INR framework for flow holographic tomography; quantitative claims outrun the validation, especially at RBC-level RI contrast. read the letter →

arxiv 2509.04848 v1 pith:ODIO5TZ4 submitted 2025-09-05 cs.CV physics.bio-phphysics.opticsq-bio.QM

Enze Ye , Wei Lin , Shaochi Ren , Yakun Liu , Xiaoping Li , Hao Wang , He Sun , Feng Pan This is my paper · ORCID
classification cs.CVphysics.bio-phphysics.opticsq-bio.QM
keywords quantitativephaseimagingflowcytometryholographictomographyimplicitneuralrepresentationrefractiveindexreconstructionposeestimationFourierdiffractiontheoremRytovapproximation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

OmniFHT claims to remove the two assumptions that currently limit in-flow holographic tomography: that each cell rotates about a single fixed axis and that the rotation angle of every projection is known in advance. Instead of detecting rotational cycles and interpolating angles, the method jointly optimizes each cell's unknown three-dimensional rotation, in-plane translation, and volumetric refractive-index distribution through a maximum-likelihood formulation in Fourier space. The volumetric map is represented by an implicit neural network evaluated on Fourier coordinates, and its built-in smoothness acts as a regularizer that fills in missing spectral information. On simulated and experimental data the paper reports high-fidelity reconstruction from as few as 10 projections or 120 degrees of angular coverage, and it reconstructs every cell in a clinical ascites specimen without pre-selection. If the claims hold, label-free flow cytometry can become unbiased with respect to cell shape and rotational dynamics.

What carries the argument

The load-bearing object is the implicit neural representation of the 3D scattering potential in Fourier space: a three-hidden-layer MLP with sinusoidal positional embedding that maps frequency coordinates $(k_x, k_y, k_z)$ to the complex scattering potential. Paired with the Fourier diffraction theorem, which linearly links a 2D Rytov-phase projection to a slice of this potential on an Ewald sphere, the network can synthesize predictions under any hypothesized pose; those predictions drive both the data-consistency loss and the pose search. The pose search is coarse-to-fine: it starts from a $30^\circ$ rotation grid and a translation grid with spacing 0.1, keeps the top eight hypotheses by complex cross-correlation, and refines by bisection over five iterations, with pose updates interleaved every five training epochs.

What would settle it

Flow a rigid phantom with known refractive-index distribution and a known rotation trajectory, such as a lithographically patterned microbead assembly on a piezo-driven rotator, through the same microfluidic channel; reconstruct with OmniFHT and compare the recovered poses and RI slices against the known ground truth. A systematic drift of RI values as scattering strength or rotation speed increases would refute the paper's implied claim that the Rytov linearization suffices for arbitrary flowing cells.

Watch

Extended reading notes

Core claim

Under the Rytov weak-scattering approximation, each measured two-dimensional complex field is a slice of the cell's three-dimensional scattering potential on an Ewald sphere, so the pose-free problem becomes a joint maximum-likelihood estimation over the unknown poses $\omega_i = (R_i, t_i)$ and the scattering potential (Eq. 5). OmniFHT solves it by alternating a coarse-to-fine search over $\mathrm{SO}(3)$ and translation space for each projection with self-supervised training of a compact multilayer perceptron that maps Fourier coordinates to the complex scattering potential; the predicted projections are generated by the Fourier diffraction theorem (Eqs. 6-7). The neural representation's implicit regularity restores missing frequencies, allowing high-fidelity reconstructions from sparse views and restricted angular ranges. The central claim is that this combination removes the single-axis, known-pose restriction of flow-based holographic tomography and enables unbiased 3D refractive-index reconstruction of entire flowing cell populations.

Load-bearing premise

The load-bearing physical premise is that each cell is a weak scatterer, so the measured complex field depends linearly on the 3D scattering potential through the Rytov approximation; if real cells or clusters scatter strongly or exhibit significant multiple scattering, the recovered refractive indices and morphologies will be biased.

Editorial extensions

If this is right

  • Cells with arbitrary geometry and multi-axis rotation, which current flow-cytometry tomography discards, can be reconstructed, removing a selection bias from population statistics.
  • Because reconstruction succeeds with as few as 10 projections or 120 degrees of angular coverage, faster flow rates and shorter in-channel dwell times become usable in high-throughput assays.
  • In simulations the method reaches 1.59 µm resolution versus 2.78 µm for the standard Rytov baseline, a 1.75-fold improvement by Fourier shell correlation.
  • On real red blood cells, multicellular aggregates, and collision-perturbed white blood cells, only OmniFHT resolves the expected morphologies (biconcave discs, distinct cell boundaries, vacuoles), where baseline reconstructions show severe artifacts.
  • In a clinical ascites specimen the method reconstructs all 21 retained cells in situ without trajectory pre-filtering, establishing a workflow for population-scale label-free RI cytometry.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the framework only requires a differentiable forward model linking structure to projections, the same joint pose-structure loop could be ported to other label-free tomographies with unknown orientation, such as X-ray or electron microscopy of isolated particles.
  • The reported resolution gains are measured against a Rytov baseline that assumes uniformly interpolated poses; comparing against a pose-free baseline given ground-truth poses would separate the benefit of pose refinement from the benefit of neural implicit regularization.
  • Swapping the Rytov model for a multiple-scattering forward model (for example a Born-series or beam-propagation operator) inside the same alternating loop is a direct testable extension that would show whether pose-free reconstruction survives outside the weak-scattering regime.
  • Since the neural representation acts as an implicit prior, it may smooth or plausibly hallucinate fine structure inside missing spectral regions; a phantom with known sub-resolution features would quantify that bias.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper introduces OmniFHT, a self-supervised framework for three-dimensional refractive-index (RI) tomography of cells flowing in a microfluidic channel. The method jointly estimates each cell's unknown 3D rotation and in-plane translation and reconstructs the 3D scattering potential as an implicit neural representation in Fourier space, using the Rytov approximation and the Fourier diffraction theorem as the forward model. The authors validate the approach on beam-propagation-method simulations of a vacuolated cell, on experimental datasets including SW780 cells, RBCs, and multicellular aggregates, and on a clinical ascites specimen containing 21 cells. They report that OmniFHT outperforms a Rytov-based baseline, recovers useful structure from as few as 5–10 projections and from 120° angular coverage, and reconstructs cells with complex multi-axis rotations that violate the single-axis assumption of conventional flow holographic tomography.

Significance. If the central claims hold, the paper would make a useful contribution: it removes the known-pose and single-axis-rotation assumptions that currently restrict in-flow holographic tomography, and it demonstrates the possibility of population-scale, label-free 3D RI imaging in clinical biofluids. The forward model and loss are clearly stated, the BPM simulation provides a ground-truth check of both pose recovery and RI reconstruction, and the experimental demonstrations on heterogeneous clinical material are valuable. The main weaknesses are that quantitative accuracy is validated only at low RI contrast, the comparison baseline is weaker than existing angle-recovery methods, and the sparse-view evaluation uses the method's own full-view output as the reference. These issues affect the strength of the quantitative and comparative claims, but they are addressable with additional validation rather than being fundamental flaws.

major comments (4)
  1. [§3.1, Eq. (3)] The quantitative-accuracy claim rests on the Rytov weak-scattering approximation, but the only ground-truth simulation uses an RI contrast of 0.02 (cell 1.35 vs. medium 1.33) and a single vacuolated cell. The target samples in Figs. 3–5 include RBCs and multicellular aggregates with substantially higher RI contrast and stronger multiple-scattering effects; in this regime the linearization in Eq. (3) can introduce RI bias and artifacts that the current BPM test does not probe. The experimental reconstructions are visually plausible but are not validated against any independent RI ground truth, so the population-scale quantitative claim is not yet supported.
  2. [§3.1, Fig. 2] The comparison baseline is described as a 'standard Rytov-based method' that assumes uniformly interpolated poses. Existing angle-recovery FHT methods (refs. 16–18) estimate the rotation from the projections rather than imposing uniform increments; because the baseline is handicapped by construction, the reported 1.75× FSC improvement and the comparisons in Figs. 2d–g and 3b–d do not establish superiority over the current state of the art. The authors should include a baseline that recovers angles from the data, such as phase-similarity matching or periodic-cycle detection, and report results for both single-axis and multi-axis rotation.
  3. [§3.2.4, Fig. 4b] The sparse-view and limited-angle FSC curves are computed against the full-view OmniFHT reconstruction of the same dataset. Because the sparse and reference reconstructions share the same INR prior, pose-initialization strategy, and optimization procedure, these FSC numbers measure consistency with the method's own full-view output rather than absolute fidelity to the true structure. The claim that reconstructions from 10 views or 120° of coverage are 'high-fidelity' should be supported by ground-truth simulations at those sampling densities, or by a reference that does not come from the same algorithm.
  4. [§2.4, Eq. (6)] The description of the neural representation is incomplete: Eq. (6) requires the network to output a complex-valued scattering potential in Fourier space, but §2.4 states that the MLP outputs a single scalar, without explaining how the real and imaginary parts are represented or how the complex loss in Eq. (7) is computed. This is central to reproducibility and should be specified (e.g., two output heads, complex-valued output layer, or separate real/imaginary channels).
minor comments (6)
  1. [§3.1, Fig. 2 caption] The caption contains a typo: 'Eular Angles' should be 'Euler Angles'; the text in §3.1 also uses 'slide' where 'slice' is intended.
  2. [§2.2, Eq. (1)] The sentence introducing Eq. (1) says 'The progress is', which appears to be a typo for 'The process is'.
  3. [§3.3] The abstract claims reconstruction of 'entire flowing cell populations', but the demonstration reconstructs 21 cells after explicitly excluding out-of-focus cells and cells near the downstream end; the abstract should be qualified to match the reported scope.
  4. [§2.3] The pose-search procedure is described only qualitatively (30° initial grid, five refinement stages, top-8 hypotheses); for reproducibility, the authors should give the exact grid sizes, translation bounds, and stopping criteria.
  5. [§3.2.4] The FSC threshold differs between the simulated (1/2, reference-based) and experimental (1/7, half-set) evaluations; the choice is plausible, but the text should state explicitly that the resolution values obtained under the two criteria are not directly comparable.
  6. [Data availability] The data availability statement is only 'Data will be made available on request'; providing code and trained models, or at least a detailed pseudocode protocol, would substantially strengthen the reproducibility of the claimed results.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the method is a self-supervised joint pose-and-structure inversion built on standard physical forward models, with independent simulation and half-set FSC checks.

full rationale

OmniFHT's derivation is self-contained and does not reduce to its inputs. The forward model in Eq. (3) is the standard Fourier diffraction theorem under the Rytov weak-scattering approximation, and Eq. (4) is the conventional relation between scattering potential and refractive index. The reconstruction loss in Eq. (7) compares the INR's predicted Rytov perturbation, synthesized via Eq. (6), directly against measured perturbations, with poses estimated by an independent coarse-to-fine search over SO(3) and R^2 grids rather than fitted to known answers. No fitted parameter is renamed as a prediction, and no load-bearing self-citation is used: the cited references to INR methods, cryo-EM reconstruction, and optical diffraction tomography are external literature, not the authors' own unverified claims. The simulation study has an explicit ground-truth cell model generated by BPM, and the experimental resolution claims use half-set FSC, which is an independent split-data consistency check. The sparse-view and limited-angle evaluations use the method's own full-view reconstruction as the FSC reference, which is a mildly self-referential evaluation metric, but that reference is not used during training or pose estimation, so it does not make the reconstruction itself circular. The Rytov weak-scattering assumption is a genuine physical limitation of the imaging model, not a circular derivation; it means quantitative accuracy at high refractive-index contrast is unverified, which is a correctness-risk concern rather than circularity. Overall, the central claim is an engineering inverse-problem result supported by its own forward model and external validation, so the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters are standard algorithmic hyperparameters chosen by hand; they affect performance but are not fitted to ground truth. The load-bearing assumptions are the Rytov approximation and the INR smoothness prior, which together enable the linear Fourier-domain model and the filling of missing spectral regions.

free parameters (6)
  • Initial SO(3) pose search interval = 30 degrees
    Hand-chosen angular step for coarse-to-fine pose search; not fitted to data, but affects convergence and accuracy.
  • Initial translation grid spacing = 0.1
    Hand-chosen spacing for translation search; affects pose estimation resolution.
  • Number of pose refinement stages = 5
    Hand-chosen number of hierarchical search rounds; more stages likely improve pose accuracy at computational cost.
  • Top-K pose hypotheses retained = 8
    Hand-chosen number of candidate poses kept between refinement stages.
  • MLP hidden layers and width = 3x256
    Hand-chosen network capacity; no ablation is shown.
  • Positional embedding octaves = 16 per coordinate
    Hand-chosen frequency range for encoding spatial coordinates.
assumptions (5)
  • domain assumption Rytov weak-scattering approximation
    Image formation model Eq. (2)-(3) assumes weak scattering so that the log-field Rytov phase is linear in the scattering potential. Valid for small RI contrast; may fail for dense aggregates or high RI.
  • standard math Fourier diffraction theorem
    Relates 2D projections to the 3D scattering potential on the Ewald sphere; cited to prior literature and used in Eq. (3).
  • domain assumption Known background refractive index nm=1.33
    Mapping from scattering potential to RI in Eq. (4) requires knowing the medium RI.
  • ad hoc to paper Smoothness/band-limited prior of INR
    The implicit neural representation implicitly regularizes the solution; the paper assumes this recovers missing spectral information, but the prior is not explicitly quantified or validated.
  • domain assumption Plane-wave illumination along z-axis
    k0x = k0y = 0 in the model; small deviations would alter the Ewald sphere sampling.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pose-Free 3D Quantitative Phase Imaging of Flowing Cellular Populations." pith.science (2026). https://pith.science/paper/ODIO5TZ4

@misc{pith2026250904848,
  author       = {Pith},
  title        = {Pith review of: Pose-Free 3D Quantitative Phase Imaging of Flowing Cellular Populations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ODIO5TZ4}},
  note         = {Machine review of arXiv:2509.04848}
}
read the original abstract

High-throughput 3D quantitative phase imaging (QPI) in flow cytometry enables label-free, volumetric characterization of individual cells by reconstructing their refractive index (RI) distributions from multiple viewing angles during flow through microfluidic channels. However, current imaging methods assume that cells undergo uniform, single-axis rotation, which require their poses to be known at each frame. This assumption restricts applicability to near-spherical cells and prevents accurate imaging of irregularly shaped cells with complex rotations. As a result, only a subset of the cellular population can be analyzed, limiting the ability of flow-based assays to perform robust statistical analysis. We introduce OmniFHT, a pose-free 3D RI reconstruction framework that leverages the Fourier diffraction theorem and implicit neural representations (INRs) for high-throughput flow cytometry tomographic imaging. By jointly optimizing each cell's unknown rotational trajectory and volumetric structure under weak scattering assumptions, OmniFHT supports arbitrary cell geometries and multi-axis rotations. Its continuous representation also allows accurate reconstruction from sparsely sampled projections and restricted angular coverage, producing high-fidelity results with as few as 10 views or only 120 degrees of angular range. OmniFHT enables, for the first time, in situ, high-throughput tomographic imaging of entire flowing cell populations, providing a scalable and unbiased solution for label-free morphometric analysis in flow cytometry platforms.

Figures

Figures reproduced from arXiv: 2509.04848 by the authors.

Figure 1
Figure 1. Overview of the OmniFHT pipeline: (a) Digital holographic flow cytometry [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Simulated validation of OmniFHT: (a) BPM-generated datasets with a 532 nm [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. OmniFHT enables high-fidelity 3D RI reconstruction of cells undergoing [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a) Left: sparse-view reconstructions of a SW780 cell in the x–y plane using [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: The figure demonstrates the 3D RI reconstructions of heterogeneous cellular [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 41 canonical work pages

  1. [1]

    Live-cell mass profiling: an emerging approach in quantitative biophysics,

    T. A. Zangle and M. A. Teitell, “Live-cell mass profiling: an emerging approach in quantitative biophysics,” Nat. methods11, 1221–1228 (2014)

    work page 2014

  2. [2]

    Quantitative phase imaging in biomedicine,

    Y. Park, C. Depeursinge, and G. Popescu, “Quantitative phase imaging in biomedicine,” Nat. photonics12, 578–589 (2018)

    work page 2018

  3. [3]

    Automatic procedure for aberration compensation in digital holographic microscopy and applications to specimen shape compensation,

    T. Colomb, E. Cuche, F. Charrière,et al., “Automatic procedure for aberration compensation in digital holographic microscopy and applications to specimen shape compensation,” Appl. optics45, 851–863 (2006)

    work page 2006

  4. [4]

    Ferraro, A

    P. Ferraro, A. Wax, and Z. Zalevsky,Coherent light microscopy: Imaging and quantitative phase analysis, vol. 46 (Springer Science & Business Media, 2011)

    work page 2011

  5. [5]

    Marker-free phase nanoscopy,

    Y. Cotte, F. Toy, P. Jourdain,et al., “Marker-free phase nanoscopy,” Nat. Photonics7, 113–117 (2013)

    work page 2013

  6. [6]

    Label-free quantitative cell division monitoring of endothelial cells by digital holographic microscopy,

    B. Kemper, A. Bauwens, A. Vollmer,et al., “Label-free quantitative cell division monitoring of endothelial cells by digital holographic microscopy,” J. Biomed. Opt.15, 036009–036009 (2010)

    work page 2010

  7. [7]

    Label-free high-resolution 3-d imaging of gold nanoparticles inside live cells using optical diffraction tomography,

    D. Kim, N. Oh, K. Kim,et al., “Label-free high-resolution 3-d imaging of gold nanoparticles inside live cells using optical diffraction tomography,” Methods136, 160–167 (2018)

    work page 2018

  8. [8]

    Living specimen tomography by digital holographic microscopy: morphometry of testate amoeba,

    F. Charrière, N. Pavillon, T. Colomb,et al., “Living specimen tomography by digital holographic microscopy: morphometry of testate amoeba,” Opt. Express14, 7005–7013 (2006)

    work page 2006

Show all 43 references
  1. [9]

    Phase contrast tomography at lab on chip scale by digital holography,

    F. Merola, P. Memmolo, L. Miccio,et al., “Phase contrast tomography at lab on chip scale by digital holography,” Methods136, 108–115 (2018)

    2018

  2. [10]

    Holotomography,

    G. Kim, H. Hugonnet, K. Kim,et al., “Holotomography,” Nat. Rev. Methods Primers4, 51 (2024)

    2024

  3. [11]

    Three-dimensional holographic refractive-index measurement of continuously flowing cells in a microfluidic channel,

    Y. Sung, N. Lue, B. Hamza,et al., “Three-dimensional holographic refractive-index measurement of continuously flowing cells in a microfluidic channel,” Phys. review applied1, 014002 (2014)

    2014

  4. [12]

    Tomographic flow cytometry by digital holography,

    F. Merola, P. Memmolo, L. Miccio,et al., “Tomographic flow cytometry by digital holography,” Light. Sci. & Appl. 6, e16241–e16241 (2017)

    2017

  5. [13]

    Three-dimensional quantitative intracellular visualization of graphene oxide nanoparticles by tomographic flow cytometry,

    D. Pirone, M. Mugnano, P. Memmolo,et al., “Three-dimensional quantitative intracellular visualization of graphene oxide nanoparticles by tomographic flow cytometry,” Nano Lett.21, 5958–5966 (2021)

    2021

  6. [14]

    3d imaging lipidometry in single cell by in-flow holographic tomography,

    D. Pirone, D. Sirico, L. Miccio,et al., “3d imaging lipidometry in single cell by in-flow holographic tomography,” Opto Electron. Adv6, 220048–220041 (2023)

    2023

  7. [15]

    Three-dimensional quantitative phase imaging of blood coagulation structures by optical projection tomography in flow cytometry using digital holographic microscopy,

    H. Funamizu and Y. Aizu, “Three-dimensional quantitative phase imaging of blood coagulation structures by optical projection tomography in flow cytometry using digital holographic microscopy,” J. biomedical optics24, 031012–031012 (2019)

    2019

  8. [16]

    Full-angle tomographic phase microscopy of flowing quasi-spherical cells,

    M. M. Villone, P. Memmolo, F. Merola,et al., “Full-angle tomographic phase microscopy of flowing quasi-spherical cells,” Lab on a Chip18, 126–131 (2018)

    2018

  9. [17]

    Rolling angle recovery of flowing cells in holographic tomography exploiting the phase similarity,

    D. Pirone, P. Memmolo, F. Merola,et al., “Rolling angle recovery of flowing cells in holographic tomography exploiting the phase similarity,” Appl. Opt.60, A277–A284 (2020)

    2020

  10. [18]

    Tomographicflowcytometrybasedonautomaticanglerecoveryforthree-dimensional imaging of carbon nanoparticles in bladder cancer cells,

    Y.Li, L.Che, W.Xiao,et al., “Tomographicflowcytometrybasedonautomaticanglerecoveryforthree-dimensional imaging of carbon nanoparticles in bladder cancer cells,” inDigital Holography and Three-Dimensional Imaging, (Optica Publishing Group, 2024), pp. Th1A–6

    2024

  11. [19]

    Occupancynetworks: Learning3dreconstructioninfunctionspace,

    L.Mescheder,M.Oechsle,M.Niemeyer,et al.,“Occupancynetworks: Learning3dreconstructioninfunctionspace,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition,(2019), pp. 4460–4470

    2019

  12. [20]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik,et al., “Nerf: Representing scenes as neural radiance fields for view synthesis,” Commun. ACM65, 99–106 (2021)

    2021

  13. [21]

    Deepsdf: Learning continuous signed distance functions for shape representation,

    J. J. Park, P. Florence, J. Straub,et al., “Deepsdf: Learning continuous signed distance functions for shape representation,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition,(2019), pp. 165–174

    2019

  14. [22]

    Implicit neural representations with periodic activation functions,

    V. Sitzmann, J. Martel, A. Bergman,et al., “Implicit neural representations with periodic activation functions,” Adv. neural information processing systems33, 7462–7473 (2020)

    2020

  15. [23]

    A structured dictionary perspective on implicit neural representations,

    G. Yüce, G. Ortiz-Jiménez, B. Besbinar, and P. Frossard, “A structured dictionary perspective on implicit neural representations,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,(2022), pp. 19228–19238

    2022

  16. [24]

    Cryodrgn: reconstruction of heterogeneous cryo-em structures using neural networks,

    E. D. Zhong, T. Bepler, B. Berger, and J. H. Davis, “Cryodrgn: reconstruction of heterogeneous cryo-em structures using neural networks,” Nat. methods18, 176–185 (2021)

    2021

  17. [25]

    Cryodrgn2: Abinitioneuralreconstructionof3dproteinstructures from real cryo-em images,

    E.D.Zhong,A.Lerer,J.H.Davis,andB.Berger,“Cryodrgn2: Abinitioneuralreconstructionof3dproteinstructures from real cryo-em images,” inProceedings of the IEEE/CVF International Conference on Computer Vision,(2021), pp. 4066–4075

    2021

  18. [26]

    Cryodrgn-et: deep reconstructing generative networks for visualizing dynamic biomolecules inside cells,

    R. Rangan, R. Feathers, S. Khavnekar,et al., “Cryodrgn-et: deep reconstructing generative networks for visualizing dynamic biomolecules inside cells,” Nat. Methods21, 1537–1545 (2024)

    2024

  19. [27]

    Dynamic ct reconstruction from limited views with implicit neural representations and parametric motion fields,

    A. W. Reed, H. Kim, R. Anirudh,et al., “Dynamic ct reconstruction from limited views with implicit neural representations and parametric motion fields,” inProceedings of the IEEE/CVF International Conference on Computer Vision,(2021), pp. 2258–2268

    2021

  20. [28]

    Nerp: implicit neural representation learning with prior embedding for sparsely sampled image reconstruction,

    L. Shen, J. Pauly, and L. Xing, “Nerp: implicit neural representation learning with prior embedding for sparsely sampled image reconstruction,” IEEE Trans. on Neural Networks Learn. Syst.35, 770–782 (2022)

    2022

  21. [29]

    Piner: Prior-informed implicit neural representation learning for test-time adaptation in sparse-view ct reconstruction,

    B. Song, L. Shen, and L. Xing, “Piner: Prior-informed implicit neural representation learning for test-time adaptation in sparse-view ct reconstruction,” inProceedings of the IEEE/CVF winter conference on applications of computer vision,(2023), pp. 1928–1938

    2023

  22. [30]

    Implicit neural representation in medical imaging: A comparative survey,

    A. Molaei, A. Aminimehr, A. Tavakoli,et al., “Implicit neural representation in medical imaging: A comparative survey,” inProceedings of the IEEE/CVF International Conference on Computer Vision,(2023), pp. 2381–2391

    2023

  23. [31]

    Eman: semiautomated software for high-resolution single-particle reconstructions,

    S. J. Ludtke, P. R. Baldwin, and W. Chiu, “Eman: semiautomated software for high-resolution single-particle reconstructions,” J. structural biology128, 82–97 (1999)

    1999

  24. [32]

    Definition and estimation of resolution in single-particle reconstructions,

    H. Y. Liao and J. Frank, “Definition and estimation of resolution in single-particle reconstructions,” Structure18, 768–775 (2010)

    2010

  25. [33]

    Missing cone artifact removal in odt using unsupervised deep learning in the projection domain,

    H. Chung, J. Huh, G. Kim,et al., “Missing cone artifact removal in odt using unsupervised deep learning in the projection domain,” IEEE Trans. on Comput. Imaging7, 747–758 (2021)

    2021

  26. [34]

    Deepregularizer: rapid resolution enhancement of tomographic imaging using deep learning,

    D. Ryu, D. Ryu, Y. Baek,et al., “Deepregularizer: rapid resolution enhancement of tomographic imaging using deep learning,” IEEE Trans. on Med. Imaging40, 1508–1518 (2021)

    2021

  27. [35]

    High-resolution three-dimensional imaging of red blood cells parasitized by plasmodium falciparum and in situ hemozoin crystals using optical diffraction tomography,

    K. Kim, H. Yoon, M. Diez-Silva,et al., “High-resolution three-dimensional imaging of red blood cells parasitized by plasmodium falciparum and in situ hemozoin crystals using optical diffraction tomography,” J. biomedical optics19, 011005–011005 (2014)

    2014

  28. [36]

    Simultaneous 3d visualization and position tracking of optically trapped particles using optical diffraction tomography,

    K. Kim, J. Yoon, and Y. Park, “Simultaneous 3d visualization and position tracking of optically trapped particles using optical diffraction tomography,” Optica2, 343–346 (2015)

    2015

  29. [37]

    Inverse problem solver for multiple light scattering using modified born series,

    M. Lee, H. Hugonnet, and Y. Park, “Inverse problem solver for multiple light scattering using modified born series,” Optica9, 177–182 (2022)

    2022

  30. [38]

    Inverse-scattering theory within the rytov approximation,

    A. Devaney, “Inverse-scattering theory within the rytov approximation,” Opt. letters6, 374–376 (1981)

    1981

  31. [39]

    Deterministic regularization of three-dimensional optical diffraction tomography,

    Y. Sung and R. R. Dasari, “Deterministic regularization of three-dimensional optical diffraction tomography,” J. Opt. Soc. Am. A28, 1554–1561 (2011)

    2011

  32. [40]

    Comparative study of iterative reconstruction algorithms for missing cone problems in optical diffraction tomography,

    J. Lim, K. Lee, K. H. Jin,et al., “Comparative study of iterative reconstruction algorithms for missing cone problems in optical diffraction tomography,” Opt. express23, 16933–16948 (2015)

    2015

  33. [41]

    Super-resolution through error energy reduction,

    R. Gerchberg, “Super-resolution through error energy reduction,” Opt. Acta: Int. J. Opt.21, 709–720 (1974)

    1974

  34. [42]

    Optical tomographic image reconstruction based on beam propagation and sparse regularization,

    U. S. Kamilov, I. N. Papadopoulos, M. H. Shoreh,et al., “Optical tomographic image reconstruction based on beam propagation and sparse regularization,” IEEE Trans. on Comput. Imaging2, 59–70 (2016)

    2016

  35. [43]

    Fourier shell correlation threshold criteria,

    M. Van Heel and M. Schatz, “Fourier shell correlation threshold criteria,” J. structural biology151, 250–262 (2005)

    2005

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.