pith. sign in

arxiv: 2606.30012 · v1 · pith:B2HHZNBZnew · submitted 2026-06-29 · 💻 cs.CV

SkelEM: Training-Signal Decoupling of Skeleton and Diffusion for Self-supervised Axial Super-Resolution in Volume Microscopy

Bohao Chen , Yanchao Zhang , Yanan Lv , Chenxun Deng , Hua Han , Xi Chen This is my paper

Pith reviewed 2026-06-30 06:42 UTC · model grok-4.3

classification 💻 cs.CV
keywords axial super-resolutionself-supervised learningdiffusion modelsskeleton extractionvolume microscopyelectron microscopymembrane segmentationzero-shot generalization
0
0 comments X

The pith

SkelEM decouples a frozen skeleton network from a diffusion refiner to enable fast, bias-free axial super-resolution from sparse microscopy slices.

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

The paper establishes that separating low-frequency topology from high-frequency detail synthesis at the training-signal level solves the trilemma of smoothed textures, hallucinations, and slow inference in self-supervised axial super-resolution. A topological network first produces a deterministic skeleton from input volumes, then a cycle-consistent mechanism extracts a real-domain residual prior from sparse slices to align and truncate the diffusion process. This yields high-fidelity outputs in five or fewer steps while preserving structures for tasks such as membrane segmentation. The approach is tested on public benchmarks and a new co-aligned dataset acquired on a Plasma-FIB instrument, showing improved generalization across modalities.

Core claim

SkelEM achieves axial super-resolution by optimizing a frozen topological network for deterministic skeletons via one objective and a diffusion refiner via a disjoint cycle-consistent objective on sparse input slices, which simultaneously extracts a real-domain residual prior and bidirectionally aligns the refiner so that the reverse diffusion process can be truncated after at most five steps without synthetic bias.

What carries the argument

The training-signal decoupling of skeleton formulation from diffusion refinement, where the skeleton supplies a deterministic low-frequency prior that enables real-domain residual extraction and early truncation of the diffusion reverse process.

If this is right

  • The method produces the most favorable fidelity-perception balance among self-supervised axial super-resolution approaches on public benchmarks.
  • SkelEM delivers state-of-the-art performance on downstream membrane segmentation tasks.
  • Zero-shot generalization holds across distinct imaging modalities without retraining.
  • Detail restoration remains high-fidelity when the diffusion process is limited to five or fewer steps.
  • The BRAVE-ASR benchmark enables rigorous measurement of cross-instrument generalization for future methods.

Where Pith is reading between the lines

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

  • The same cycle-consistent residual extraction could be adapted to other inverse problems where only anisotropic acquisitions are available.
  • Truncating diffusion at five steps suggests the skeleton prior captures most of the necessary structural information, which may reduce compute demands in high-throughput volume imaging pipelines.
  • If the topological network can be replaced by other deterministic structure extractors, the framework might extend to non-microscopy domains that suffer from directional resolution limits.

Load-bearing premise

A frozen topological network produces a deterministic skeleton that can be used to extract a real-domain residual prior and truncate the reverse diffusion process without introducing synthetic bias or structural hallucinations.

What would settle it

A head-to-head comparison on the BRAVE-ASR benchmark in which SkelEM produces lower membrane segmentation accuracy or more visible structural hallucinations than either pure regression or full-step diffusion baselines would falsify the benefit of the decoupling.

Figures

Figures reproduced from arXiv: 2606.30012 by Bohao Chen, Chenxun Deng, Hua Han, Xi Chen, Yanan Lv, Yanchao Zhang.

Figure 1
Figure 1. Figure 1: SkelEM demonstrates robust axial super-resolution performance across differ￾ent imaging modalities and instruments. Top row (yellow box): A challenging 8× axial super-resolution zero-shot instrument transfer task on our BRAVE-ASR dataset. Bottom row (blue box): A real-world (no high-resolution reference) 10× axial super￾resolution task on a volume light microscopy dataset. 1 Introduction Volume microscopy … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the SkelEM framework. Top: Three-stage training. (Left) fflow is trained on Vˆhigh to produce a smooth structural skeleton; fdetail is deliberately dis￾carded to enforce topology–texture separation. (Middle) The frozen fflow cyclically reconstructs observed slices from Vlow, and a residual estimator f∆ is trained via fre￾quency loss to enhance the high-frequency details. (Right) The pretrained … view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of SkelEM with state-of-the-art self-supervised ASR methods on the EPFL dataset (r = 8) across two orthogonal views (XY and XZ). For each view, we show the reconstructed slices with zoomed-in detail insets, fre￾quency spectra, and intensity profiles (orange) compared against the ground-truth (blue dashed), alongside the corresponding 3D membrane segmentation results. Red arrows highl… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the BRAVE-ASR dataset. Left: the isotropic reference volume (3708×3000×800 voxels at 5×5×5 nm3 ). Right: three co-aligned physically acquired anisotropic volumes (2000×2000×256 voxels each) at slice thicknesses of 10 nm, 20 nm, and 40 nm, corresponding to 2×, 4×, and 8× axial anisotropy. All volumes share the same 5 nm lateral pixel size and were acquired from mouse brain tissue using a Thermo … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of degradation strategies on BRAVE-ASR [PITH_FULL_IMAGE:figures/full_fig_p024_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Quantitative Resolution Comparison on BRAVE-ASR. [PITH_FULL_IMAGE:figures/full_fig_p024_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Tri-view (XY, XZ, YZ) qualitative comparison on the FIB-25 dataset (8× ASR). Color-coded insets and red arrows highlight structural differences. Supervised methods are marked with (Sup.). τ of XY slice is 0.5 [PITH_FULL_IMAGE:figures/full_fig_p029_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Tri-view (XY, XZ, YZ) qualitative comparison on the EPFL dataset (8× ASR). Color-coded insets and red arrows highlight structural differences. Supervised methods are marked with (Sup.). τ of XY slice is 0.5 [PITH_FULL_IMAGE:figures/full_fig_p030_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Normalized rank heatmap across all evaluated metrics. [PITH_FULL_IMAGE:figures/full_fig_p032_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Limitation analysis on resolving fine double-membrane structures. Yellow ar￾rows indicate the membrane gap. SkelEM exhibits blurring in the gap region. In con￾trast, InterpolAI, leveraging strong object permanence priors from natural video train￾ing, enforces a clearer geometric separation but fails to synthesize biologically realistic textures, producing scene-constant interpolations that lack structural… view at source ↗
read the original abstract

Volume microscopy, including electron and light microscopy, suffers from severe anisotropic resolution due to physical axial sectioning. Existing self-supervised axial super-resolution (ASR) methods face a trilemma bounded by overly smoothed regression textures, structural hallucinations of pure diffusion models, and prohibitive inference latency. In this paper, we propose Skeleton-refinE Microscopy (SkelEM), a self-supervised framework that decouples ASR at the training-signal level: a frozen topological network and a diffusion refiner are optimized by disjoint objectives, separating low-frequency topology formulation from high-frequency detail enhancement. Building on this deterministic skeleton, we exploit a unified cycle-consistent mechanism on input sparse slices to simultaneously extract a real-domain residual prior and bidirectionally align the diffusion refiner, washing away cross-plane artifacts without synthetic bias. By truncating the reverse diffusion process with this physical prior, SkelEM achieves high-fidelity detail restoration in merely $\le 5$ steps. To rigorously assess cross-instrument generalization, we further introduce BRAVE-ASR, a new benchmark of co-aligned anisotropic and isotropic volumes acquired on a Plasma-FIB instrument. Across public benchmarks, SkelEM achieves the most favorable balance across the fidelity-perception trade-off among self-supervised methods, with state-of-the-art downstream membrane segmentation performance and robust zero-shot generalization across distinct modalities.

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 proposes SkelEM, a self-supervised framework for axial super-resolution in volume microscopy. It decouples the training signal by using a frozen topological network to extract a deterministic skeleton and a diffusion refiner for high-frequency details. A cycle-consistent mechanism on input sparse slices extracts a real-domain residual prior and aligns the refiner bidirectionally, allowing truncation of the reverse diffusion process to ≤5 steps without synthetic bias. The method claims the most favorable fidelity-perception trade-off among self-supervised methods, SOTA downstream membrane segmentation, and robust zero-shot generalization on public benchmarks and the new BRAVE-ASR benchmark.

Significance. If the central claims hold, SkelEM would offer an efficient solution to the trilemma in self-supervised ASR by balancing fidelity and perception while avoiding hallucinations and high latency, with strong performance in downstream tasks and cross-modality generalization. This could have significant impact in volume microscopy applications.

major comments (2)
  1. [Methods (cycle-consistent mechanism and skeleton extraction)] The central claim relies on the frozen topological network producing a deterministic skeleton from anisotropic sparse slices that enables bias-free residual prior extraction and safe truncation of diffusion at ≤5 steps. However, no explicit validation is provided that the skeleton remains topologically faithful on real low-SNR axial data, nor that the bidirectional alignment eliminates rather than regularizes cross-plane artifacts. This is load-bearing for the 'no synthetic bias' guarantee.
  2. [Abstract and Results] The abstract states favorable trade-offs and SOTA segmentation performance but provides no quantitative metrics, ablation results, or error analysis to support these claims, making verification of the balance across fidelity-perception trade-off difficult.
minor comments (1)
  1. [Notation and Methods] Clarify the definition of the residual prior and how it is extracted from the cycle-consistent mechanism to avoid ambiguity in the truncation step.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment point by point below and indicate planned revisions.

read point-by-point responses

  1. Referee: [Methods (cycle-consistent mechanism and skeleton extraction)] The central claim relies on the frozen topological network producing a deterministic skeleton from anisotropic sparse slices that enables bias-free residual prior extraction and safe truncation of diffusion at ≤5 steps. However, no explicit validation is provided that the skeleton remains topologically faithful on real low-SNR axial data, nor that the bidirectional alignment eliminates rather than regularizes cross-plane artifacts. This is load-bearing for the 'no synthetic bias' guarantee.

    Authors: We agree that direct validation of topological fidelity on low-SNR axial data would strengthen the central claim. In the revised manuscript we will add quantitative evaluation of skeleton accuracy (e.g., topological error metrics and structure-preservation scores) on held-out low-SNR slices from both public datasets and BRAVE-ASR. We will also include an ablation isolating the bidirectional cycle-consistent alignment to demonstrate that it reduces cross-plane artifacts beyond simple regularization, with supporting error analysis. revision: yes

  2. Referee: [Abstract and Results] The abstract states favorable trade-offs and SOTA segmentation performance but provides no quantitative metrics, ablation results, or error analysis to support these claims, making verification of the balance across fidelity-perception trade-off difficult.

    Authors: The abstract is a high-level summary constrained by length limits. We will revise it to incorporate the key quantitative results already present in the main text (fidelity-perception scores, segmentation accuracies, and latency comparisons) so that the claimed trade-offs are directly supported by numbers. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation separates a frozen topological network (producing deterministic skeleton via disjoint objective) from a diffusion refiner, then applies cycle-consistency on input sparse slices to extract residual prior for truncation. No equations, self-citations, or ansatzes are exhibited that reduce any claimed prediction or prior to a fitted input or self-definition by construction. The cycle-consistent extraction is presented as operating on the given anisotropic slices to remove cross-plane artifacts, with the topological network held fixed and objectives explicitly disjoint; this structure is independent of the target super-resolution output. External elements such as the new BRAVE-ASR benchmark and downstream segmentation metrics further anchor the claims outside any internal fit.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the central claim rests on the unstated assumption that disjoint objectives on skeleton and diffusion produce a usable physical prior without further fitting.

pith-pipeline@v0.9.1-grok · 5786 in / 1228 out tokens · 26818 ms · 2026-06-30T06:42:58.483884+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

53 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Aurelien, L., Yunpeng, L., Carlos, B., Pascal, F.:https://www.epfl.ch/labs/ cvlab/data/data-em/(2013), accessed: 2024-08-16

    2013

  2. [2]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Blau, Y., Michaeli, T.: The perception-distortion tradeoff. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 6228–6237 (2018)

    2018

  3. [3]

    Journal of the Peripheral Nervous System30(2), e70019 (2025)

    Borisovs, V., Bossi, M., Matino, L., Marmiroli, P., Cavaletti, G.: New approaches based on serial-block face electron microscopy to investigate the peripheral nervous system. Journal of the Peripheral Nervous System30(2), e70019 (2025)

    2025

  4. [4]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Chen, B., Zhang, Y., Lv, Y., Han, H., Chen, X.: Self-supervised axial super- resolution for volume microscopy via diffusion-guided structure distillation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 467–477. Springer (2025)

    2025

  5. [5]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Chen, Q., Chen, X., Wang, C., Liu, Y., Xiong, Z., Wu, F.: Learning multimodal volumetric features for large-scale neuron tracing. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 1174–1182 (2024) 16 B. Chen et al

    2024

  6. [6]

    In: Med- ical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part V 23

    Deng, S., Fu, X., Xiong, Z., Chen, C., Liu, D., Chen, X., Ling, Q., Wu, F.: Isotropic reconstruction of 3d em images with unsupervised degradation learning. In: Med- ical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part V 23. pp. 163–173. Springer (2020)

    2020

  7. [7]

    Advances in neural information processing systems34, 8780–8794 (2021)

    Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. Advances in neural information processing systems34, 8780–8794 (2021)

    2021

  8. [8]

    Current pro- tocols in cytometry92(1), e68 (2020)

    Elliott, A.D.: Confocal microscopy: principles and modern practices. Current pro- tocols in cytometry92(1), e68 (2020)

    2020

  9. [9]

    IEEE transactions on image processing17(10), 1737–1754 (2008)

    Foi, A., Trimeche, M., Katkovnik, V., Egiazarian, K.: Practical poissonian-gaussian noise modeling and fitting for single-image raw-data. IEEE transactions on image processing17(10), 1737–1754 (2008)

    2008

  10. [10]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Hai, Y., Wang, G., Su, T., Jiang, W., Hu, Y.: Hierarchical flow diffusion for ef- ficient frame interpolation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 22943–22952 (2025)

    2025

  11. [11]

    In: Computer vision: a reference guide, pp

    Hasinoff, S.W.: Photon, poisson noise. In: Computer vision: a reference guide, pp. 980–982. Springer (2021)

    2021

  12. [12]

    bioRxiv pp

    He, J., Zhang, Y., Sun, W., Yang, G., Sun, F.: Isovem: Isotropic reconstruction for volume electron microscopy based on transformer. bioRxiv pp. 2023–11 (2023)

    2023

  13. [13]

    In: Medical Image Computing and Computer-Assisted Intervention- MICCAI 2017: 20th International Conference, Quebec City, QC, Canada, September 11-13, 2017, Proceedings, Part II 20

    Heinrich,L.,Bogovic,J.A.,Saalfeld,S.:Deeplearningforisotropicsuper-resolution from non-isotropic 3d electron microscopy. In: Medical Image Computing and Computer-Assisted Intervention- MICCAI 2017: 20th International Conference, Quebec City, QC, Canada, September 11-13, 2017, Proceedings, Part II 20. pp. 135–143. Springer (2017)

    2017

  14. [14]

    Advances in neural information processing systems30(2017)

    Heusel,M.,Ramsauer,H.,Unterthiner,T.,Nessler,B.,Hochreiter,S.:Ganstrained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems30(2017)

    2017

  15. [15]

    Advances in neural information processing systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

    2020

  16. [16]

    In: European Conference on Computer Vision

    Huang, Z., Zhang, T., Heng, W., Shi, B., Zhou, S.: Real-time intermediate flow estimation for video frame interpolation. In: European Conference on Computer Vision. pp. 624–642. Springer (2022)

    2022

  17. [17]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Jain, S., Watson, D., Tabellion, E., Poole, B., Kontkanen, J., et al.: Video inter- polation with diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7341–7351 (2024)

    2024

  18. [18]

    In: Proceedings of the Computer Vision and Pat- tern Recognition Conference

    Jeong, M.W., Rhee, C.E.: Lc-mamba: Local and continuous mamba with shifted windows for frame interpolation. In: Proceedings of the Computer Vision and Pat- tern Recognition Conference. pp. 17671–17681 (2025)

    2025

  19. [19]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Jiang, C., Gedeon, A., Lyu, Y., Landgraf, E., Zhang, Y., Hou, X., Kondepudi, A., Chowdury, A., Lee, H., Hollon, T.: Super-resolution of biomedical volumes with 2d supervision. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 6966–6977 (2024)

    2024

  20. [20]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Jiang, H., Sun, D., Jampani, V., Yang, M.H., Learned-Miller, E., Kautz, J.: Super slomo: High quality estimation of multiple intermediate frames for video interpo- lation. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 9000–9008 (2018)

    2018

  21. [21]

    Nature Methods pp

    Joshi, S., Forjaz, A., Han, K.S., Shen, Y., Queiroga, V., Selaru, F.A., Gérard, M., Xenes, D., Matelsky, J., Wester, B., et al.: Interpolai: deep learning-based optical flow interpolation and restoration of biomedical images for improved 3d tissue mapping. Nature Methods pp. 1–12 (2025) SkelEM: Training-Signal Decoupling for Volume Microscopy ASR 17

    2025

  22. [22]

    Journal of Mi- croscopy287(3), 114–137 (2022)

    Kievits, A.J., Lane, R., Carroll, E.C., Hoogenboom, J.P.: How innovations in methodology offer new prospects for volume electron microscopy. Journal of Mi- croscopy287(3), 114–137 (2022)

    2022

  23. [23]

    Accessed June 29, 2026

    Koho, S., Kalinin, A.: MIPLIB: A Python software library for microscopy image restoration, reconstruction and analysis.https://github.com/sakoho81/miplib (2019), version v1.0. Accessed June 29, 2026

    2019

  24. [24]

    Nature communications10(1), 3103 (2019)

    Koho, S., Tortarolo, G., Castello, M., Deguchi, T., Diaspro, A., Vicidomini, G.: Fourier ring correlation simplifies image restoration in fluorescence microscopy. Nature communications10(1), 3103 (2019)

    2019

  25. [25]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Lee, H.J., Jo, E., Lim, M., Son, Y.H., Kang, B., Nam, H., Jeong, J.H., Shin, D.H., Kam, T.E.: Sparse3diff: A diffusion framework for 3d reconstruction from sparse 2d slices in volumetric optical imaging. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 510–519. Springer (2025)

    2025

  26. [26]

    In: International Conference on Medical Image Computing and Computer- Assisted Intervention

    Lee, K., Jeong, W.K.: Reference-free isotropic 3d em reconstruction using diffusion models. In: International Conference on Medical Image Computing and Computer- Assisted Intervention. pp. 235–245. Springer (2023)

    2023

  27. [27]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Lee,K.,Jeong,W.K.:Reference-freeaxialsuper-resolutionof3dmicroscopyimages using implicit neural representation with a 2d diffusion prior. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 593–602. Springer (2024)

    2024

  28. [28]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Lee, S., Chung, H., Park, M., Park, J., Ryu, W.S., Ye, J.C.: Improving 3d imag- ing with pre-trained perpendicular 2d diffusion models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 10710–10720 (2023)

    2023

  29. [29]

    Journal of neuroscience methods226, 88–102 (2014)

    Liu, T., Jones, C., Seyedhosseini, M., Tasdizen, T.: A modular hierarchical ap- proach to 3d electron microscopy image segmentation. Journal of neuroscience methods226, 88–102 (2014)

    2014

  30. [30]

    In: Computer Vision–ECCV 2020 Workshops: Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16

    Liu, Y., Xie, L., Siyao, L., Sun, W., Qiao, Y., Dong, C.: Enhanced quadratic video interpolation. In: Computer Vision–ECCV 2020 Workshops: Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16. pp. 41–56. Springer (2020)

    2020

  31. [31]

    Bioinformatics 38(24), 5329–5339 (2022)

    Liu, Y., Wang, G., Ascoli, G.A., Zhou, J., Liu, L.: Neuron tracing from light microscopy images: automation, deep learning and bench testing. Bioinformatics 38(24), 5329–5339 (2022)

    2022

  32. [32]

    Nature Communications15(1), 4677 (2024)

    Lu, C., Chen, K., Qiu, H., Chen, X., Chen, G., Qi, X., Jiang, H.: Diffusion-based deep learning method for augmenting ultrastructural imaging and volume electron microscopy. Nature Communications15(1), 4677 (2024)

    2024

  33. [33]

    Biology11(9), 1270 (2022)

    Ma, H., Chen, J., Deng, Z., Sun, T., Luo, Q., Gong, H., Li, X., Long, B.: Multi- scale analysis of cellular composition and morphology in intact cerebral organoids. Biology11(9), 1270 (2022)

    2022

  34. [34]

    Journal of mul- tivariate analysis98(5), 873–895 (2007)

    Meilă, M.: Comparing clusterings—an information based distance. Journal of mul- tivariate analysis98(5), 873–895 (2007)

    2007

  35. [35]

    SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations

    Meng, C., He, Y., Song, Y., Song, J., Wu, J., Zhu, J.Y., Ermon, S.: Sdedit: Guided image synthesis and editing with stochastic differential equations. arXiv preprint arXiv:2108.01073 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  36. [36]

    Auto- matica11(285-296), 23–27 (1975)

    Otsu, N., et al.: A threshold selection method from gray-level histograms. Auto- matica11(285-296), 23–27 (1975)

    1975

  37. [37]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Pan, M., Gan, Y., Zhou, F., Liu, J., Zhang, Y., Wang, A., Zhang, S., Li, D.: Diffuseir: Diffusion models for isotropic reconstruction of 3d microscopic images. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 323–332. Springer (2023) 18 B. Chen et al

    2023

  38. [38]

    Nature Communications13(1), 3297 (2022)

    Park, H., Na, M., Kim, B., Park, S., Kim, K.H., Chang, S., Ye, J.C.: Deep learn- ing enables reference-free isotropic super-resolution for volumetric fluorescence mi- croscopy. Nature Communications13(1), 3297 (2022)

    2022

  39. [39]

    Nature Reviews Methods Primers2(1), 51 (2022)

    Peddie, C.J., Genoud, C., Kreshuk, A., Meechan, K., Micheva, K.D., Narayan, K., Pape, C., Parton, R.G., Schieber, N.L., Schwab, Y., et al.: Volume electron microscopy. Nature Reviews Methods Primers2(1), 51 (2022)

    2022

  40. [40]

    Schwartz, J., Jiang, Y., Wang, Y., Aiello, A., Bhattacharya, P., Yuan, H., Mi, Z., Bassim, N., Hovden, R.: Removing stripes, scratches, and curtaining with nonre- coverablecompressedsensing.MicroscopyandMicroanalysis25(3),705–710(2019)

    2019

  41. [41]

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models (2020)

    2020

  42. [42]

    Sun, H., Ye, E., Lu, C., Jiang, Z., Bai, W., Ren, S., Wang, C., Zhang, J., Shi, R., Ma, L., et al.: Em generalist: A physics-driven diffusion foundation model for electron microscopy (2025)

    2025

  43. [43]

    Proceedings of the National Academy of Sciences112(44), 13711–13716 (2015)

    Takemura, S.y., Xu, C.S., Lu, Z., Rivlin, P.K., Parag, T., Olbris, D.J., Plaza, S., Zhao, T., Katz, W.T., Umayam, L., et al.: Synaptic circuits and their variations within different columns in the visual system of drosophila. Proceedings of the National Academy of Sciences112(44), 13711–13716 (2015)

    2015

  44. [44]

    MICCAI Workshop on Efficient Medical AI (EMA) (2025).https: //doi.org/10.1101/2024.09.07.611785

    Troidl, J., Liang, Y., Beyer, J., Tavakoli, M., Danzl, J.G., Hadwiger, M., Pfister, H., Tompkin, J.: niiv: Interactive self-supervised neural implicit isotropic volume reconstruction. MICCAI Workshop on Efficient Medical AI (EMA) (2025).https: //doi.org/10.1101/2024.09.07.611785

    work page doi:10.1101/2024.09.07.611785 2025

  45. [45]

    In: Medical Image Computing and Computer-Assisted Intervention - MICCAI 2017

    Weigert, M., Royer, L., Jug, F., Myers, G.: Isotropic reconstruction of 3d fluores- cence microscopy images using convolutional neural networks. In: Medical Image Computing and Computer-Assisted Intervention - MICCAI 2017. pp. 126–134. Springer International Publishing, Cham (2017)

    2017

  46. [46]

    Nature methods15(12), 1090–1097 (2018)

    Weigert, M., Schmidt, U., Boothe, T., Müller, A., Dibrov, A., Jain, A., Wilhelm, B., Schmidt, D., Broaddus, C., Culley, S., et al.: Content-aware image restoration: pushing the limits of fluorescence microscopy. Nature methods15(12), 1090–1097 (2018)

    2018

  47. [47]

    Plos one17(12), e0279005 (2022)

    Wu, S., Nakao, M., Imanishi, K., Nakamura, M., Mizowaki, T., Matsuda, T.: Com- puted tomography slice interpolation in the longitudinal direction based on deep learning techniques: To reduce slice thickness or slice increment without dose in- crease. Plos one17(12), e0279005 (2022)

    2022

  48. [48]

    elife6, e25916 (2017)

    Xu, C.S., Hayworth, K.J., Lu, Z., Grob, P., Hassan, A.M., García-Cerdán, J.G., Niyogi, K.K., Nogales, E., Weinberg, R.J., Hess, H.F.: Enhanced fib-sem systems for large-volume 3d imaging. elife6, e25916 (2017)

    2017

  49. [49]

    Biomedical Optics Express11(11), 6634–6648 (2020)

    Ye, S., Yin, Y., Yao, J., Nie, J., Song, Y., Gao, Y., Yu, J., Li, H., Fei, P., Zheng, W.: Axial resolution improvement of two-photon microscopy by multi-frame re- construction and adaptive optics. Biomedical Optics Express11(11), 6634–6648 (2020)

    2020

  50. [50]

    In: Proceedings of the Computer Vision and Pattern Recognition Con- ference

    Yue, Z., Liao, K., Loy, C.C.: Arbitrary-steps image super-resolution via diffusion inversion. In: Proceedings of the Computer Vision and Pattern Recognition Con- ference. pp. 23153–23163 (2025)

    2025

  51. [51]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 586–595 (2018)

    2018

  52. [52]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Zhang, Y., Guo, J., Zhai, H., Liu, J., Han, H.: Segneuron: 3d neuron instance seg- mentation in any em volume with a generalist model. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 589–600. Springer (2024),https://github.com/yanchaoz/SegNeuron SkelEM: Training-Signal Decoupling for Volume Microscopy ASR 19

    2024

  53. [53]

    In: 2019 international conference on 3D vision (3DV)

    Zhou, D., Fang, J., Song, X., Guan, C., Yin, J., Dai, Y., Yang, R.: Iou loss for 2d/3d object detection. In: 2019 international conference on 3D vision (3DV). pp. 85–94. IEEE (2019) 20 B. Chen et al. A Supplementary Materials Overview Thissupplementarymaterialprovidescomprehensivetheoreticalanalyses,imple- mentation details, and extensive experimental eva...

    work page doi:10.5281/zenodo.18920195 2019