SimAQ: Mitigating Experimental Artifacts in Soft X-Ray Tomography using Simulated Acquisitions

Daniel W\"ustner; Jacob Egebjerg

arxiv: 2508.10821 · v3 · pith:LJEY7FBEnew · submitted 2025-08-14 · 🧬 q-bio.QM

SimAQ: Mitigating Experimental Artifacts in Soft X-Ray Tomography using Simulated Acquisitions

Jacob Egebjerg , Daniel W\"ustner This is my paper

Pith reviewed 2026-05-21 22:42 UTC · model grok-4.3

classification 🧬 q-bio.QM

keywords soft x-ray tomographysimulation pipelineyeast phantomsneural network segmentationtransfer learningexperimental artifactsmissing wedgefew-shot learning

0 comments

The pith

A simulation pipeline creates synthetic yeast tomograms with realistic artifacts to train neural networks that segment real experimental data using few or no labels.

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

The paper develops SimAQ as a way to generate large volumes of paired synthetic data consisting of yeast phantoms, sinograms, and reconstructions that include common experimental issues such as the missing wedge. Networks trained mostly or entirely on this synthetic data are shown to transfer to real soft X-ray tomograms and produce usable segmentations. This matters because real annotated datasets for whole-cell tomography remain small and expensive to create, which currently limits quantitative measurements of cellular structure in noisy images. If the transfer works, researchers could perform reliable analysis on existing experimental volumes without first collecting and labeling thousands of real examples.

Core claim

SimAQ generates realistic yeast phantoms and applies synthetic imaging artifacts to produce paired noisy volumes, sinograms, and reconstructions; neural networks trained primarily on this synthetic data achieve accurate segmentations on real soft X-ray tomograms through effective few-shot and zero-shot transfer learning.

What carries the argument

The SimAQ simulation pipeline that builds yeast phantoms and overlays experimental artifacts to create training pairs for segmentation models.

If this is right

Quantitative segmentation of cellular features becomes feasible on noisy real tomograms without requiring large manually labeled real datasets.
Few-shot and zero-shot transfer from synthetic to real data reduces the annotation burden for new experiments.
Paired synthetic sinograms and reconstructions allow direct training of models that handle missing-wedge artifacts.
The same pipeline can supply unlimited training examples once the phantom generation rules are set.

Where Pith is reading between the lines

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

The approach could be adapted to other cell types or imaging modalities that suffer from similar limited-angle artifacts.
Pre-training on simulation might serve as a general strategy for scientific imaging tasks where real ground-truth labels are scarce.
Combining the synthetic data with a small number of real examples could further improve robustness to variations in experimental conditions.

Load-bearing premise

The simulated yeast phantoms and added artifacts are statistically and structurally close enough to real experimental soft X-ray data that models trained on them will transfer successfully.

What would settle it

Training a segmentation network on SimAQ data and then testing it on a held-out collection of real tomograms acquired with different experimental parameters or cell types would show whether the transfer accuracy drops below usable levels.

Figures

Figures reproduced from arXiv: 2508.10821 by Daniel W\"ustner, Jacob Egebjerg.

**Figure 2.** Figure 2: Projections from a real tomogram of a yeast cell acquired at Bessy II (left) and from [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Training scheme: The encoder encodes the noisy reconstructed phantom, and the decoder [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Examples of synthetic slices used for evaluation. Top row: noisy input reconstructions. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗

**Figure 5.** Figure 5: Violin plot of IoU scores for 3,000 slices across different cell counts and channels. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗

**Figure 6.** Figure 6: Rendered raw predictions on real tomograms. Top row: 3D renderings of raw model [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

**Figure 7.** Figure 7: Effect of simple post-processing on two real yeast cell tomograms. Left panels (a) show [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗

read the original abstract

Soft X-ray tomography provides detailed structural insight into whole cells but is hindered by experimental artifacts such as the missing wedge and by limited availability of annotated datasets. We present SimAQ, a simulation pipeline that generates realistic yeast phantoms and applies synthetic imaging artifacts to produce paired noisy volumes, sinograms, and reconstructions. We validate our approach by training a neural network primarily on synthetic data and demonstrate effective few-shot and zero-shot transfer learning on real X-ray tomograms. Our model delivers accurate segmentations, enabling quantitative analysis of noisy tomograms without relying on large labeled datasets.

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.

Desk Editor's Note private letter to a colleague

SimAQ's simulation pipeline for yeast soft X-ray tomography enables data-efficient segmentation but transfer performance rests on unverified sim-to-real fidelity.

read the letter

The punchline here is that SimAQ generates simulated yeast phantoms and applies synthetic artifacts to create training data for neural networks, allowing them to segment real soft X-ray tomograms with minimal real labels through few-shot and zero-shot learning. What stands out as new is the tailored pipeline for this modality and cell type. Previous work has used simulations for tomography in other contexts, but the details for soft X-ray of yeast, including how they model the missing wedge and reconstruction artifacts, add a concrete tool. The paper does well by focusing on a clear bottleneck in the field: not enough annotated real data for training reliable models. By showing transfer works to some degree, it points to a way forward for quantitative analysis of cell structures in noisy tomograms. The soft spots come down to validation. The central assumption is that the simulated data is close enough to real experiments for the network to generalize. If the yeast phantoms don't include enough variation in organelle placement or membrane details, or if the artifact simulation doesn't match the actual noise and missing information patterns in experiments, then the reported accurate segmentations might not translate to new datasets. The abstract doesn't include numbers on how they measured this fidelity or compared against baselines, so it's difficult to judge the robustness without those details in the results section. This paper is for researchers in soft X-ray cellular imaging who are limited by data annotation costs. A reader looking for methods to bootstrap models from simulation would get direct value from the described approach and could adapt the phantoms for their own work. It deserves serious peer review. The idea addresses a genuine practical issue and the simulation framework is a step in the right direction, even if more evidence on the sim-to-real gap would strengthen it.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces SimAQ, a simulation pipeline that generates realistic yeast phantoms and applies synthetic soft X-ray tomography artifacts (missing wedge, noise, reconstruction effects) to produce paired noisy volumes, sinograms, and reconstructions. The central claim is that a neural network trained primarily on this synthetic data achieves effective few-shot and zero-shot transfer learning to real experimental tomograms, yielding accurate segmentations that enable quantitative analysis without large labeled real datasets.

Significance. If the simulation fidelity is sufficient to support the observed transfer, the work would address a key bottleneck in soft X-ray tomography by reducing reliance on scarce annotated real data and providing a route to artifact-robust segmentation. This could facilitate broader quantitative cellular studies in a modality constrained by experimental limitations.

major comments (2)

[Validation and Results] Validation/Results: The claim that the model 'delivers accurate segmentations' on real tomograms via synthetic-to-real transfer is not supported by any reported quantitative metrics (e.g., Dice scores, IoU, Hausdorff distance), baseline comparisons (real-data-only training or alternative artifact-correction methods), error bars, or details on data exclusion/validation splits. Without these, the transfer performance cannot be assessed as evidence of robustness rather than simulation-specific artifacts.
[Methods] Methods, phantom and artifact generation: The central transfer-learning claim rests on the unquantified assumption that the yeast phantoms plus synthetic missing-wedge, noise, and reconstruction artifacts produce a distribution sufficiently close to real soft X-ray tomograms. No quantitative fidelity checks (power-spectrum comparison, spatial autocorrelation, intensity histogram overlap, or structural similarity metrics between synthetic and experimental volumes) are described to confirm this overlap.

minor comments (2)

[Abstract] Abstract: The description of 'few-shot and zero-shot transfer learning' would benefit from explicit numbers of real samples used in the few-shot regime and clarification of the zero-shot protocol.
[Figures and Notation] Notation and figures: Ensure consistent use of terms such as 'sinogram' versus 'projection' and improve clarity of any figures showing synthetic versus real tomogram comparisons.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which have helped us strengthen the manuscript. We have revised the paper to include the requested quantitative metrics for both transfer performance and simulation fidelity, providing stronger evidence for the synthetic-to-real transfer claims.

read point-by-point responses

Referee: [Validation and Results] Validation/Results: The claim that the model 'delivers accurate segmentations' on real tomograms via synthetic-to-real transfer is not supported by any reported quantitative metrics (e.g., Dice scores, IoU, Hausdorff distance), baseline comparisons (real-data-only training or alternative artifact-correction methods), error bars, or details on data exclusion/validation splits. Without these, the transfer performance cannot be assessed as evidence of robustness rather than simulation-specific artifacts.

Authors: We agree that quantitative metrics are necessary to rigorously support the transfer-learning results. In the revised manuscript we now report Dice scores, IoU, and Hausdorff distances on the real tomograms, together with comparisons against a real-data-only training baseline and an alternative artifact-correction approach. Error bars from repeated training runs and explicit descriptions of data splits and exclusion criteria have been added to the Results and Methods sections. revision: yes
Referee: [Methods] Methods, phantom and artifact generation: The central transfer-learning claim rests on the unquantified assumption that the yeast phantoms plus synthetic missing-wedge, noise, and reconstruction artifacts produce a distribution sufficiently close to real soft X-ray tomograms. No quantitative fidelity checks (power-spectrum comparison, spatial autocorrelation, intensity histogram overlap, or structural similarity metrics between synthetic and experimental volumes) are described to confirm this overlap.

Authors: We acknowledge that explicit quantitative fidelity checks strengthen the justification for the observed transfer. The revised manuscript now includes power-spectrum comparisons, intensity-histogram overlap statistics, spatial-autocorrelation analysis, and SSIM values between synthetic and experimental volumes. These results are presented in a new supplementary figure and accompanying text to demonstrate distributional similarity. revision: yes

Circularity Check

0 steps flagged

No circularity: external real-data validation anchors the transfer claim

full rationale

The paper constructs synthetic yeast phantoms and applies simulated artifacts (missing wedge, noise, reconstruction effects) to train a segmentation network, then measures performance via few-shot and zero-shot transfer on independent real soft X-ray tomograms. This constitutes an external empirical benchmark rather than any derivation that reduces to the simulation inputs by construction. No equations, fitted parameters renamed as predictions, or load-bearing self-citations are described that would close a loop; the central claim rests on observed accuracy against real experimental data outside the synthetic distribution.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that synthetic data faithfully reproduces real experimental characteristics; no free parameters or invented entities are mentioned in the abstract.

axioms (1)

domain assumption Synthetic imaging artifacts applied to simulated phantoms accurately represent real experimental artifacts such as the missing wedge in soft X-ray tomography.
This assumption underpins the generation of realistic training data that enables transfer to real tomograms.

pith-pipeline@v0.9.0 · 5621 in / 1197 out tokens · 51679 ms · 2026-05-21T22:42:37.547349+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Yeast phantoms are simulated using ellipsoids with radii sampled from Gaussian distributions... Ice cracks are added using pre-generated 128×128×128 volumes modulated by low-frequency Perlin noise... angular range sampled from uniform [70°,130°]... FBP with Ram-Lak and Hamming filter.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We validate our approach by training a neural network primarily on synthetic data and demonstrate effective few-shot and zero-shot transfer learning on real X-ray tomograms.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · 1 internal anchor

[1]

Soft X-ray tomography and cryogenic light microscopy: the cool combination in cellular imaging

Gerry McDermott et al. “Soft X-ray tomography and cryogenic light microscopy: the cool combination in cellular imaging”. In: Trends in cell biology 19.11 (2009), pp. 587–595

work page 2009
[2]

Soft X-ray tomography: virtual sculptures from cell cultures

Jessica Guo and Carolyn A Larabell. “Soft X-ray tomography: virtual sculptures from cell cultures”. In: Current opinion in structural biology 58 (2019), pp. 324–332

work page 2019
[3]

Cryo-soft X-ray tomography: a journey into the world of the native-state cell

Raffaella Carzaniga et al. “Cryo-soft X-ray tomography: a journey into the world of the native-state cell”. In: Protoplasma 251 (2014), pp. 449–458

work page 2014
[4]

Cryo X-ray microscope with flat sample geometry for correlative fluorescence and nanoscale tomographic imaging

Gerd Schneider et al. “Cryo X-ray microscope with flat sample geometry for correlative fluorescence and nanoscale tomographic imaging”. In: Journal of structural biology 177.2 (2012), pp. 212–223

work page 2012
[5]

Soft X-ray tomograms provide a structural basis for whole-cell modeling

Valentina Loconte et al. “Soft X-ray tomograms provide a structural basis for whole-cell modeling”. In: The FASEB Journal37.1 (2023), e22681

work page 2023
[6]

Soft X-Ray Tomography Has Evolved into a Pow- erful Tool for Revealing Cell Structures

Venera Weinhardt and Carolyn Larabell. “Soft X-Ray Tomography Has Evolved into a Pow- erful Tool for Revealing Cell Structures”. In: Annual Review of Analytical Chemistry 18 (2025)

work page 2025
[7]

The ill-conditioned nature of the limited angle tomography problem

Mark E Davison. “The ill-conditioned nature of the limited angle tomography problem”. In: SIAM Journal on Applied Mathematics 43.2 (1983), pp. 428–448

work page 1983
[8]

Characterization and reduction of artifacts in limited angle tomography

Jürgen Frikel and Eric Todd Quinto. “Characterization and reduction of artifacts in limited angle tomography”. In: Inverse Problems 29.12 (2013), p. 125007

work page 2013
[9]

Radiographic techniques, contrast, and noise in x-ray imaging

Walter Huda and R Brad Abrahams. “Radiographic techniques, contrast, and noise in x-ray imaging”. In: American Journal of Roentgenology 204.2 (2015), W126–W131

work page 2015
[10]

Low-dose CT reconstruction via edge-preserving total variation regulariza- tion

Zhen Tian et al. “Low-dose CT reconstruction via edge-preserving total variation regulariza- tion”. In: Physics in Medicine & Biology 56.18 (2011), p. 5949

work page 2011
[11]

Unsupervised Learnable Sinogram Inpainting Network (SIN) for Limited Angle CT reconstruction

Ji Zhao et al. “Unsupervised learnable sinogram inpainting network (SIN) for limited angle CT reconstruction”. In: arXiv preprint arXiv:1811.03911 (2018). 13

work page internal anchor Pith review Pith/arXiv arXiv 2018
[12]

Directional sinogram inpainting for limited angle tomography

Robert Tovey et al. “Directional sinogram inpainting for limited angle tomography”. In:Inverse problems 35.2 (2019), p. 024004

work page 2019
[13]

Dictionary learning based sinogram inpainting for CT sparse reconstruction

Si Li et al. “Dictionary learning based sinogram inpainting for CT sparse reconstruction”. In: Optik 125.12 (2014), pp. 2862–2867

work page 2014
[14]

0.7 Å resolution electron tomography enabled by deep-learning-aided information recovery

Chunyang Wang et al. “0.7 Å resolution electron tomography enabled by deep-learning-aided information recovery”. In: Advanced Intelligent Systems 2.12 (2020), p. 2000152

work page 2020
[15]

Isotropic reconstruction for electron tomography with deep learning

Yun-Tao Liu et al. “Isotropic reconstruction for electron tomography with deep learning”. In: Nature communications 13.1 (2022), p. 6482

work page 2022
[16]

Three-dimensional cellular ultrastructure resolved by X-ray microscopy

Gerd Schneider et al. “Three-dimensional cellular ultrastructure resolved by X-ray microscopy”. In: Nature methods 7.12 (2010), pp. 985–987

work page 2010
[17]

Computed tomography of cryogenic biological specimens based on X-ray microscopic images

D Weiss et al. “Computed tomography of cryogenic biological specimens based on X-ray microscopic images”. In: Ultramicroscopy 84.3-4 (2000), pp. 185–197

work page 2000
[18]

Quantitative analysis of yeast internal architecture using soft X-ray tomography

Maho Uchida et al. “Quantitative analysis of yeast internal architecture using soft X-ray tomography”. In: Yeast 28.3 (2011), pp. 227–236

work page 2011
[19]

Visualizing subcellular rearrangements in intact β cells using soft x-ray tomography

Kate L White et al. “Visualizing subcellular rearrangements in intact β cells using soft x-ray tomography”. In: Science advances 6.50 (2020), eabc8262

work page 2020
[20]

Quantitative imaging of membrane contact sites for sterol transfer between endo-lysosomes and mitochondria in living cells

Alice Dupont Juhl et al. “Quantitative imaging of membrane contact sites for sterol transfer between endo-lysosomes and mitochondria in living cells”. In: Scientific reports 11.1 (2021), p. 8927

work page 2021
[21]

Soft X-ray tomography to map and quantify organelle interactions at the mesoscale

Valentina Loconte et al. “Soft X-ray tomography to map and quantify organelle interactions at the mesoscale”. In: Structure 30.4 (2022), pp. 510–521

work page 2022
[22]

Automated quantification of vacuole fusion and lipophagy in Saccharomyces cerevisiae from fluorescence and cryo-soft X-ray microscopy data using deep learning

Jacob Marcus Egebjerg et al. “Automated quantification of vacuole fusion and lipophagy in Saccharomyces cerevisiae from fluorescence and cryo-soft X-ray microscopy data using deep learning”. In: Autophagy 20.4 (2024), pp. 902–922

work page 2024
[23]

Structural changes in cells imaged by soft X-ray cryo- tomography during hepatitis C virus infection

Ana Joaquina Pérez-Berná et al. “Structural changes in cells imaged by soft X-ray cryo- tomography during hepatitis C virus infection”. In: ACS nano 10.7 (2016), pp. 6597–6611

work page 2016
[24]

3D correlative cryo-structured illumination fluorescence and soft X-ray microscopy elucidates reovirus intracellular release pathway

Ilias Kounatidis et al. “3D correlative cryo-structured illumination fluorescence and soft X-ray microscopy elucidates reovirus intracellular release pathway”. In: Cell 182.2 (2020), pp. 515– 530

work page 2020
[25]

Imaging of virus-infected cells with soft X-ray tomography

Damià Garriga et al. “Imaging of virus-infected cells with soft X-ray tomography”. In: Viruses 13.11 (2021), p. 2109

work page 2021
[26]

Using soft X-ray tomography for rapid whole-cell quantitative imaging of SARS-CoV-2-infected cells

Valentina Loconte et al. “Using soft X-ray tomography for rapid whole-cell quantitative imaging of SARS-CoV-2-infected cells”. In:Cell reports methods 1.7 (2021)

work page 2021
[27]

Near-native state imaging by cryo-soft-X-ray tomography reveals remodelling of multiple cellular organelles during HSV-1 infection

Kamal L Nahas et al. “Near-native state imaging by cryo-soft-X-ray tomography reveals remodelling of multiple cellular organelles during HSV-1 infection”. In:PLoS Pathogens 18.7 (2022), e1010629

work page 2022
[28]

Cells undergo major changes in the quantity of cytoplasmic organelles after uptake of gold nanoparticles with biologically relevant surface coatings

Burcu Kepsutlu et al. “Cells undergo major changes in the quantity of cytoplasmic organelles after uptake of gold nanoparticles with biologically relevant surface coatings”. In:ACS nano 14.2 (2020), pp. 2248–2264

work page 2020
[29]

Niemann Pick C2 protein enables cholesterol transfer from endo- lysosomes to the plasma membrane for efflux by shedding of extracellular vesicles

Alice Dupont Juhl et al. “Niemann Pick C2 protein enables cholesterol transfer from endo- lysosomes to the plasma membrane for efflux by shedding of extracellular vesicles”. In: Chemistry and Physics of Lipids 235 (2021), p. 105047

work page 2021
[30]

Ergosterol promotes aggregation of natamycin in the yeast plasma mem- brane

Maria Szomek et al. “Ergosterol promotes aggregation of natamycin in the yeast plasma mem- brane”. In: Biochimica et Biophysica Acta (BBA)-Biomembranes 1866.7 (2024), p. 184350

work page 2024
[31]

Bayesian approach to limited-angle reconstruc- tion in computed tomography

Kenneth M Hanson and George W Wecksung. “Bayesian approach to limited-angle reconstruc- tion in computed tomography”. In: Journal of the Optical Society of America 73.11 (1983), pp. 1501–1509

work page 1983
[32]

3D PSF Measurement for a Soft X-ray Microscope and Comparison to Theory

JG McNally et al. “3D PSF Measurement for a Soft X-ray Microscope and Comparison to Theory”. In: Computational Optical Sensing and Imaging. Optica Publishing Group. 2016, pp. CM3D–4

work page 2016
[33]

Tomo3D 2.0–exploitation of advanced vector extensions (A VX) for 3D reconstruction

Jose-Ignacio Agulleiro and Jose-Jesus Fernandez. “Tomo3D 2.0–exploitation of advanced vector extensions (A VX) for 3D reconstruction”. In:Journal of structural biology189.2 (2015), pp. 147–152

work page 2015
[34]

No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges

Lehan Yao et al. “No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges”. In: npj Computational Materials 10.1 (2024), p. 28. 14

work page 2024
[35]

Automated segmentation of soft X-ray tomography: native cellular structure with sub-micron resolution at high throughput for whole-cell quantitative imaging in yeast

Jianhua Chen et al. “Automated segmentation of soft X-ray tomography: native cellular structure with sub-micron resolution at high throughput for whole-cell quantitative imaging in yeast”. In: bioRxiv (2024), pp. 2024–10

work page 2024
[36]

Automated 3D cytoplasm segmentation in soft X-ray tomography

Ayse Erozan et al. “Automated 3D cytoplasm segmentation in soft X-ray tomography”. In: Iscience 27.6 (2024)

work page 2024
[37]

3D surface reconstruction of cellular cryo-soft X-ray microscopy tomograms using semisupervised deep learning

Michael CA Dyhr et al. “3D surface reconstruction of cellular cryo-soft X-ray microscopy tomograms using semisupervised deep learning”. In: Proceedings of the National Academy of Sciences 120.24 (2023), e2209938120

work page 2023
[38]

Simulating the cellular context in synthetic datasets for cryo-electron tomography

Antonio Martinez-Sanchez et al. “Simulating the cellular context in synthetic datasets for cryo-electron tomography”. In: IEEE Transactions on Medical Imaging (2024)

work page 2024
[39]

The Apollonian packing of circles

Edward Kasner and Fred Supnick. “The Apollonian packing of circles”. In:Proceedings of the National Academy of Sciences 29.11 (1943), pp. 378–384

work page 1943
[40]

An image synthesizer

Ken Perlin. “An image synthesizer”. In: ACM Siggraph Computer Graphics 19.3 (1985), pp. 287–296

work page 1985
[41]

TorchRadon: Fast Differentiable Routines for Computed Tomography

Matteo Ronchetti. “TorchRadon: Fast Differentiable Routines for Computed Tomography”. In: arXiv preprint arXiv:2009.14788 (2020). eprint: arXiv:2009.14788

work page arXiv 2009
[42]

Shortcut learning in deep neural networks

Robert Geirhos et al. “Shortcut learning in deep neural networks”. In: Nature Machine Intelli- gence 2.11 (2020), pp. 665–673

work page 2020
[43]

Unpaired image-to-image translation using cycle-consistent adversarial networks

Jun-Yan Zhu et al. “Unpaired image-to-image translation using cycle-consistent adversarial networks”. In: Proceedings of the IEEE international conference on computer vision. 2017, pp. 2223–2232

work page 2017
[44]

Tomographic reconstruction in soft x-ray microscopy using focus-stack back-projection

Mårten Selin et al. “Tomographic reconstruction in soft x-ray microscopy using focus-stack back-projection”. In: Optics letters 40.10 (2015), pp. 2201–2204. Appendix A Hardware We train the model used in Section 4 on DeiC Interactive HPC and evaluate running times on our local workstation. Hardware specifications Consumer Hardware Operating System Windows...

work page 2015

[1] [1]

Soft X-ray tomography and cryogenic light microscopy: the cool combination in cellular imaging

Gerry McDermott et al. “Soft X-ray tomography and cryogenic light microscopy: the cool combination in cellular imaging”. In: Trends in cell biology 19.11 (2009), pp. 587–595

work page 2009

[2] [2]

Soft X-ray tomography: virtual sculptures from cell cultures

Jessica Guo and Carolyn A Larabell. “Soft X-ray tomography: virtual sculptures from cell cultures”. In: Current opinion in structural biology 58 (2019), pp. 324–332

work page 2019

[3] [3]

Cryo-soft X-ray tomography: a journey into the world of the native-state cell

Raffaella Carzaniga et al. “Cryo-soft X-ray tomography: a journey into the world of the native-state cell”. In: Protoplasma 251 (2014), pp. 449–458

work page 2014

[4] [4]

Cryo X-ray microscope with flat sample geometry for correlative fluorescence and nanoscale tomographic imaging

Gerd Schneider et al. “Cryo X-ray microscope with flat sample geometry for correlative fluorescence and nanoscale tomographic imaging”. In: Journal of structural biology 177.2 (2012), pp. 212–223

work page 2012

[5] [5]

Soft X-ray tomograms provide a structural basis for whole-cell modeling

Valentina Loconte et al. “Soft X-ray tomograms provide a structural basis for whole-cell modeling”. In: The FASEB Journal37.1 (2023), e22681

work page 2023

[6] [6]

Soft X-Ray Tomography Has Evolved into a Pow- erful Tool for Revealing Cell Structures

Venera Weinhardt and Carolyn Larabell. “Soft X-Ray Tomography Has Evolved into a Pow- erful Tool for Revealing Cell Structures”. In: Annual Review of Analytical Chemistry 18 (2025)

work page 2025

[7] [7]

The ill-conditioned nature of the limited angle tomography problem

Mark E Davison. “The ill-conditioned nature of the limited angle tomography problem”. In: SIAM Journal on Applied Mathematics 43.2 (1983), pp. 428–448

work page 1983

[8] [8]

Characterization and reduction of artifacts in limited angle tomography

Jürgen Frikel and Eric Todd Quinto. “Characterization and reduction of artifacts in limited angle tomography”. In: Inverse Problems 29.12 (2013), p. 125007

work page 2013

[9] [9]

Radiographic techniques, contrast, and noise in x-ray imaging

Walter Huda and R Brad Abrahams. “Radiographic techniques, contrast, and noise in x-ray imaging”. In: American Journal of Roentgenology 204.2 (2015), W126–W131

work page 2015

[10] [10]

Low-dose CT reconstruction via edge-preserving total variation regulariza- tion

Zhen Tian et al. “Low-dose CT reconstruction via edge-preserving total variation regulariza- tion”. In: Physics in Medicine & Biology 56.18 (2011), p. 5949

work page 2011

[11] [11]

Unsupervised Learnable Sinogram Inpainting Network (SIN) for Limited Angle CT reconstruction

Ji Zhao et al. “Unsupervised learnable sinogram inpainting network (SIN) for limited angle CT reconstruction”. In: arXiv preprint arXiv:1811.03911 (2018). 13

work page internal anchor Pith review Pith/arXiv arXiv 2018

[12] [12]

Directional sinogram inpainting for limited angle tomography

Robert Tovey et al. “Directional sinogram inpainting for limited angle tomography”. In:Inverse problems 35.2 (2019), p. 024004

work page 2019

[13] [13]

Dictionary learning based sinogram inpainting for CT sparse reconstruction

Si Li et al. “Dictionary learning based sinogram inpainting for CT sparse reconstruction”. In: Optik 125.12 (2014), pp. 2862–2867

work page 2014

[14] [14]

0.7 Å resolution electron tomography enabled by deep-learning-aided information recovery

Chunyang Wang et al. “0.7 Å resolution electron tomography enabled by deep-learning-aided information recovery”. In: Advanced Intelligent Systems 2.12 (2020), p. 2000152

work page 2020

[15] [15]

Isotropic reconstruction for electron tomography with deep learning

Yun-Tao Liu et al. “Isotropic reconstruction for electron tomography with deep learning”. In: Nature communications 13.1 (2022), p. 6482

work page 2022

[16] [16]

Three-dimensional cellular ultrastructure resolved by X-ray microscopy

Gerd Schneider et al. “Three-dimensional cellular ultrastructure resolved by X-ray microscopy”. In: Nature methods 7.12 (2010), pp. 985–987

work page 2010

[17] [17]

Computed tomography of cryogenic biological specimens based on X-ray microscopic images

D Weiss et al. “Computed tomography of cryogenic biological specimens based on X-ray microscopic images”. In: Ultramicroscopy 84.3-4 (2000), pp. 185–197

work page 2000

[18] [18]

Quantitative analysis of yeast internal architecture using soft X-ray tomography

Maho Uchida et al. “Quantitative analysis of yeast internal architecture using soft X-ray tomography”. In: Yeast 28.3 (2011), pp. 227–236

work page 2011

[19] [19]

Visualizing subcellular rearrangements in intact β cells using soft x-ray tomography

Kate L White et al. “Visualizing subcellular rearrangements in intact β cells using soft x-ray tomography”. In: Science advances 6.50 (2020), eabc8262

work page 2020

[20] [20]

Quantitative imaging of membrane contact sites for sterol transfer between endo-lysosomes and mitochondria in living cells

Alice Dupont Juhl et al. “Quantitative imaging of membrane contact sites for sterol transfer between endo-lysosomes and mitochondria in living cells”. In: Scientific reports 11.1 (2021), p. 8927

work page 2021

[21] [21]

Soft X-ray tomography to map and quantify organelle interactions at the mesoscale

Valentina Loconte et al. “Soft X-ray tomography to map and quantify organelle interactions at the mesoscale”. In: Structure 30.4 (2022), pp. 510–521

work page 2022

[22] [22]

Automated quantification of vacuole fusion and lipophagy in Saccharomyces cerevisiae from fluorescence and cryo-soft X-ray microscopy data using deep learning

Jacob Marcus Egebjerg et al. “Automated quantification of vacuole fusion and lipophagy in Saccharomyces cerevisiae from fluorescence and cryo-soft X-ray microscopy data using deep learning”. In: Autophagy 20.4 (2024), pp. 902–922

work page 2024

[23] [23]

Structural changes in cells imaged by soft X-ray cryo- tomography during hepatitis C virus infection

Ana Joaquina Pérez-Berná et al. “Structural changes in cells imaged by soft X-ray cryo- tomography during hepatitis C virus infection”. In: ACS nano 10.7 (2016), pp. 6597–6611

work page 2016

[24] [24]

3D correlative cryo-structured illumination fluorescence and soft X-ray microscopy elucidates reovirus intracellular release pathway

Ilias Kounatidis et al. “3D correlative cryo-structured illumination fluorescence and soft X-ray microscopy elucidates reovirus intracellular release pathway”. In: Cell 182.2 (2020), pp. 515– 530

work page 2020

[25] [25]

Imaging of virus-infected cells with soft X-ray tomography

Damià Garriga et al. “Imaging of virus-infected cells with soft X-ray tomography”. In: Viruses 13.11 (2021), p. 2109

work page 2021

[26] [26]

Using soft X-ray tomography for rapid whole-cell quantitative imaging of SARS-CoV-2-infected cells

Valentina Loconte et al. “Using soft X-ray tomography for rapid whole-cell quantitative imaging of SARS-CoV-2-infected cells”. In:Cell reports methods 1.7 (2021)

work page 2021

[27] [27]

Near-native state imaging by cryo-soft-X-ray tomography reveals remodelling of multiple cellular organelles during HSV-1 infection

Kamal L Nahas et al. “Near-native state imaging by cryo-soft-X-ray tomography reveals remodelling of multiple cellular organelles during HSV-1 infection”. In:PLoS Pathogens 18.7 (2022), e1010629

work page 2022

[28] [28]

Cells undergo major changes in the quantity of cytoplasmic organelles after uptake of gold nanoparticles with biologically relevant surface coatings

Burcu Kepsutlu et al. “Cells undergo major changes in the quantity of cytoplasmic organelles after uptake of gold nanoparticles with biologically relevant surface coatings”. In:ACS nano 14.2 (2020), pp. 2248–2264

work page 2020

[29] [29]

Niemann Pick C2 protein enables cholesterol transfer from endo- lysosomes to the plasma membrane for efflux by shedding of extracellular vesicles

Alice Dupont Juhl et al. “Niemann Pick C2 protein enables cholesterol transfer from endo- lysosomes to the plasma membrane for efflux by shedding of extracellular vesicles”. In: Chemistry and Physics of Lipids 235 (2021), p. 105047

work page 2021

[30] [30]

Ergosterol promotes aggregation of natamycin in the yeast plasma mem- brane

Maria Szomek et al. “Ergosterol promotes aggregation of natamycin in the yeast plasma mem- brane”. In: Biochimica et Biophysica Acta (BBA)-Biomembranes 1866.7 (2024), p. 184350

work page 2024

[31] [31]

Bayesian approach to limited-angle reconstruc- tion in computed tomography

Kenneth M Hanson and George W Wecksung. “Bayesian approach to limited-angle reconstruc- tion in computed tomography”. In: Journal of the Optical Society of America 73.11 (1983), pp. 1501–1509

work page 1983

[32] [32]

3D PSF Measurement for a Soft X-ray Microscope and Comparison to Theory

JG McNally et al. “3D PSF Measurement for a Soft X-ray Microscope and Comparison to Theory”. In: Computational Optical Sensing and Imaging. Optica Publishing Group. 2016, pp. CM3D–4

work page 2016

[33] [33]

Tomo3D 2.0–exploitation of advanced vector extensions (A VX) for 3D reconstruction

Jose-Ignacio Agulleiro and Jose-Jesus Fernandez. “Tomo3D 2.0–exploitation of advanced vector extensions (A VX) for 3D reconstruction”. In:Journal of structural biology189.2 (2015), pp. 147–152

work page 2015

[34] [34]

No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges

Lehan Yao et al. “No ground truth needed: unsupervised sinogram inpainting for nanoparticle electron tomography (UsiNet) to correct missing wedges”. In: npj Computational Materials 10.1 (2024), p. 28. 14

work page 2024

[35] [35]

Automated segmentation of soft X-ray tomography: native cellular structure with sub-micron resolution at high throughput for whole-cell quantitative imaging in yeast

Jianhua Chen et al. “Automated segmentation of soft X-ray tomography: native cellular structure with sub-micron resolution at high throughput for whole-cell quantitative imaging in yeast”. In: bioRxiv (2024), pp. 2024–10

work page 2024

[36] [36]

Automated 3D cytoplasm segmentation in soft X-ray tomography

Ayse Erozan et al. “Automated 3D cytoplasm segmentation in soft X-ray tomography”. In: Iscience 27.6 (2024)

work page 2024

[37] [37]

3D surface reconstruction of cellular cryo-soft X-ray microscopy tomograms using semisupervised deep learning

Michael CA Dyhr et al. “3D surface reconstruction of cellular cryo-soft X-ray microscopy tomograms using semisupervised deep learning”. In: Proceedings of the National Academy of Sciences 120.24 (2023), e2209938120

work page 2023

[38] [38]

Simulating the cellular context in synthetic datasets for cryo-electron tomography

Antonio Martinez-Sanchez et al. “Simulating the cellular context in synthetic datasets for cryo-electron tomography”. In: IEEE Transactions on Medical Imaging (2024)

work page 2024

[39] [39]

The Apollonian packing of circles

Edward Kasner and Fred Supnick. “The Apollonian packing of circles”. In:Proceedings of the National Academy of Sciences 29.11 (1943), pp. 378–384

work page 1943

[40] [40]

An image synthesizer

Ken Perlin. “An image synthesizer”. In: ACM Siggraph Computer Graphics 19.3 (1985), pp. 287–296

work page 1985

[41] [41]

TorchRadon: Fast Differentiable Routines for Computed Tomography

Matteo Ronchetti. “TorchRadon: Fast Differentiable Routines for Computed Tomography”. In: arXiv preprint arXiv:2009.14788 (2020). eprint: arXiv:2009.14788

work page arXiv 2009

[42] [42]

Shortcut learning in deep neural networks

Robert Geirhos et al. “Shortcut learning in deep neural networks”. In: Nature Machine Intelli- gence 2.11 (2020), pp. 665–673

work page 2020

[43] [43]

Unpaired image-to-image translation using cycle-consistent adversarial networks

Jun-Yan Zhu et al. “Unpaired image-to-image translation using cycle-consistent adversarial networks”. In: Proceedings of the IEEE international conference on computer vision. 2017, pp. 2223–2232

work page 2017

[44] [44]

Tomographic reconstruction in soft x-ray microscopy using focus-stack back-projection

Mårten Selin et al. “Tomographic reconstruction in soft x-ray microscopy using focus-stack back-projection”. In: Optics letters 40.10 (2015), pp. 2201–2204. Appendix A Hardware We train the model used in Section 4 on DeiC Interactive HPC and evaluate running times on our local workstation. Hardware specifications Consumer Hardware Operating System Windows...

work page 2015