Pith. sign in

REVIEW 3 major objections 5 minor 51 references

CRIS: Cross-Plane Self-Supervised Isotropic Restoration for Anisotropic Volumetric Imaging Across Modalities

T0 review · 3 major / 5 minor · reviewed 2026-07-12 · grok-4.5

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

Pith's one-line read CRIS restores isotropic 3D volumes from anisotropic MRI and electron-microscopy stacks without paired ground truth by completing missing stripes on two orthogonal views and fusing them.

desk verdict Solid methods paper: blank-slice cross-plane self-supervision with real multi-modality numbers and a useful single-model gap/plane robustness result; main soft spot is synthetic-degradation transfer, not the core idea. read the letter →

arxiv 2606.15967 v2 pith:K6DRTNIT submitted 2026-06-14 cs.CV

Adi Ahituv , Anat Ilivitzki , Moti Freiman This is my paper
classification cs.CV
keywords isotropicrestorationanisotropicvolumesself-supervisedlearningsuper-resolutionMRIvolumeelectronmicroscopycross-planefusionstripecompletion
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

Thick-slice MRI and sectioned volume electron microscopy leave the through-plane axis sparse, so orthogonal reformats look staircased and segmentation or tracing suffer. CRIS shows that the high-resolution in-plane slices already present in the same volume can supervise a 2D completion network: those slices are synthetically degraded and periodically masked so the network learns to fill the missing stripes. At test time blank slices define an isotropic grid, two orthogonal reformats are restored independently, and their predictions are averaged. On brain MRI the method reaches about 33 dB PSNR and 0.96 SSIM, improves downstream Dice over strong baselines, and a single model trained with variable gap factors stays above interpolation from 3 imes to 7 imes anisotropy in every plane. Parallel gains hold on abdominal MRI (reference-free FID/KID) and on two vEM benchmarks up to 8 imes, including noisy hemibrain data. The practical claim is that one cross-plane self-supervised recipe, with only modality-specific degradation operators, removes the need for isotropic targets or per-configuration retraining.

What carries the argument

Cross-plane stripe completion: blank-slice insertion turns the 3D problem into 2D periodic-mask completion on two orthogonal reformats; a Swin-UNet-style network trained only on degraded in-plane slices restores each reformat, and voxel-wise averaging fuses them into a consensus isotropic volume.

What would settle it

Train CRIS on truly anisotropic clinical or microscopy stacks that have matched isotropic re-acquisitions of the same subjects, then measure whether its PSNR/SSIM/Dice gains over SMORE-style and neural-implicit baselines disappear or reverse relative to the synthetic-anisotropy numbers reported in the paper.

Watch

Extended reading notes

Core claim

Isotropic restoration from a single anisotropic volume can be solved as self-supervised 2D stripe completion on orthogonal reformats of a blank-padded isotropic grid, without ever seeing paired isotropic ground truth; multi-view averaging of the two restored reformats yields volumes that outperform interpolation and recent self-supervised baselines on both MRI and volume electron microscopy, and one variable-gap model generalizes across anisotropy factors and planes.

Load-bearing premise

The method assumes that synthetic degradations—directional Gaussian blur for MRI and average pooling for electron microscopy—plus in-plane self-supervision are close enough to real through-plane physics that performance on synthetic public volumes and reference-free clinical scans will transfer to everyday acquisitions.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. CRIS reformulates isotropic restoration of anisotropic volumes as self-supervised 2D stripe completion on orthogonal reformats of a blank-padded isotropic grid, without paired isotropic ground truth. High-resolution in-plane slices are modality-specifically degraded (directional 1D Gaussian blur for MRI; average pooling for vEM) and periodically masked for training; at inference, blank slices define the target grid, two orthogonal reformats are restored by a Swin-UNet-style network, and predictions are fused by voxel-wise averaging. The method is evaluated on public multi-site brain MRI (g=5, plus a single variable-gap model across gaps 3–7 and three planes), private reference-free abdominal FIESTA MRI, and EPFL/hemibrain vEM up to 8× anisotropy, with comparisons to interpolation and strong self-supervised baselines (SMORE4, SIMPLE, SA-INR, ATME, NIIV, vEMINR), plus SynthSeg segmentation consistency and component ablations.

Significance. If the results transfer, CRIS is a practically useful, modality-flexible route to isotropic restoration without paired isotropic targets or per-configuration retraining—an important setting for clinical MRI and large-scale vEM. Strengths include a clear blank-slice vs. interpolation-refinement distinction from SMORE-style methods, multi-cohort evaluation with case-level means±SD and Wilcoxon+Holm tests, 32-structure SynthSeg consistency, a component ablation (Table 4), a single-model variable-gap robustness study (Fig. 4, Table S3), and public code. The cross-modality framing with shared stripe-completion and modality-aware degradations is a coherent contribution beyond single-domain self-supervised SR.

major comments (3)
  1. The modality-flexible / real-acquisition claim is only weakly supported for true through-plane physics. Public brain MRI and both vEM benchmarks use synthetic anisotropy (Sec. 3.3, 4.1–4.3: σ=0.75√g Gaussian + every-g-th retention; average pooling). Real MRI slice profiles and vEM sectioning (knife marks, compression, staining) are not matched. The only real-anisotropy cohort (abdominal FIESTA, Table 5) is reference-free and judged only by FID/KID vs. native in-plane slices, which cannot establish through-plane anatomical fidelity. Discussion acknowledges this but still asserts transfer. Please either (i) add a real thick-slice evaluation with an independent isotropic or multi-plane reference where possible, or (ii) substantially qualify the abstract/conclusion claims to synthetic-anisotropy and reference-free perceptual settings, and state what would be required to claim clinical/vEM tr
  2. Baseline fairness is uneven and load-bearing for the “outperforms SMORE4/SIMPLE/SA-INR/ATME/NIIV/vEMINR” claim (Tables 1, 5–8). SA-INR is adapted without paired isotropic targets and trained per test case; SIMPLE is applied cross-domain on brain MRI; ATME uses interpolated inputs with plane-specific adversarial training; SMORE4 is per-volume internal learning. Supplementary baseline notes help, but the main text should state more clearly which methods were re-tuned under matched self-supervised constraints and which were off-domain or handicapped. Without that, ranking gaps (e.g., SIMPLE/SA-INR far below interpolation on brain MRI) are hard to interpret as pure method superiority rather than protocol mismatch.
  3. The single-model robustness experiment (Sec. 5.2, Fig. 4, Table S3) is valuable but compared only to interpolation, while the primary claim is superiority to learning-based methods without configuration-specific retraining. At minimum, report one learning baseline (e.g., SMORE4 or a fixed-g CRIS) under the same cross-gap/cross-plane protocol, or explicitly limit the claim to “generalizes without retraining relative to interpolation and fixed-g training,” rather than implying broad learning-based robustness.
minor comments (5)
  1. Abstract vs. body metric inconsistencies: abstract reports 32.921/0.963 and EPFL 29.100/0.830 and 26.874/0.722, while Table 1 and Tables 6–7 report slightly different SSIM/PSNR (e.g., 0.963±0.003 vs. 0.9631±0.0027; EPFL 29.133/0.834 and 27.123/0.734). Align all reported numbers.
  2. Abstract lists ECLARE among outperformed methods, but ECLARE does not appear in Tables 1 or 5 or the baseline sections. Remove or add the comparison.
  3. Fig. 1 and Sec. 3.6: clarify that the two “orthogonal reformats” are generated from one padded volume (not two acquired scans), as stated later in Sec. 3.2; this is easy to misread from the overview figure alone.
  4. Table 2: single-view vs. fused trade-offs are interesting; a short sentence on when a user should prefer single-view (view-matched perception) vs. fused (volumetric metrics/downstream) would help practitioners.
  5. Minor notation/typo cleanup: SSIM reported as 0.9631 in one place and 0.963 elsewhere; ensure consistent rounding and that “Edges 3D” is defined once in the main text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical self-supervised restoration evaluated on held-out references, not a derivation that defines outputs from fitted inputs.

full rationale

CRIS is an empirical deep-learning method, not a first-principles derivation. Training constructs self-supervised pairs from observed high-resolution in-plane slices via modality-specific degradation and periodic masking (Eqs. 1–6); the network learns stripe completion, and inference inserts blanks, restores two orthogonal reformats, and averages them (Eq. 12). Reported PSNR/SSIM/Dice/FID are measured against held-out isotropic test volumes (brain MRI, EPFL, hemibrain) or native in-plane comparators (abdominal MRI), not against quantities that were fitted into the model. Hyperparameters (σ≈0.75√g, loss weights, gap sets) are selected on validation data—standard practice, not a fitted-input-called-prediction of the test metrics. Citations to SMORE, NIIV, vEMINR, and co-authored SIMPLE are baselines or related work; none supply a uniqueness theorem or ansatz that forces the central claim. Concerns about synthetic degradation fidelity affect transfer/correctness risk, not circularity of the derivation chain.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

Load-bearing content is mostly engineering and domain modeling, not new physics. Free parameters are standard ML/hyperparameter choices (gap set, blur σ schedule, loss weights, architecture size, microscopy contrast α). Axioms are domain assumptions that synthetic degradations and in-plane self-supervision suffice. Invented entities are methodological constructs (CRIS formulation, insertion operator), not new physical objects.

free parameters (5)
  • MRI slice-profile σ schedule (σ = 0.75√g)
    Chosen via validation FID/PSNR calibration (Fig. 2) rather than measured per-scan slice profiles; primary brain results use g=5 → σ≈1.67.
  • Training gap factor set
    Fixed g=5 for main brain benchmark; g∈{6,7} for abdominal; g∈{3..7} for robustness model; separate 4×/8× runs for EPFL—selected to match cohort spacing, not derived.
  • Composite loss weights (α,β,γ,δ) and staged activation epochs
    MRI uses α=10, β=5, γ=5, δ=10 with SSIM/FFL/Sobel staged from epochs 3/5/10; vEM sets γ=δ=0 and down-weights SSIM—hand-tuned schedules.
  • Microscopy contrast-restoration coefficient α
    Percentile-based reverse-degradation correction P̂HR = PLR + α(PLR − PLLR) with α selected on validation (Sec. 3.7).
  • Network/training hyperparameters (patch size, LR, batch, base filters)
    Dataset-specific values in Table S2 (e.g., 256² patches and 1e-4 LR for brain MRI) chosen for optimization stability, not predicted by theory.
assumptions (4)
  • domain assumption Through-plane information loss can be adequately simulated by directional 1D Gaussian blur (MRI) or average pooling (vEM) plus periodic masking of high-resolution in-plane slices.
    Sec. 3.3 and Related Work; central to constructing self-supervised pairs without isotropic GT.
  • domain assumption High-resolution in-plane structure provides enough conditional information to reconstruct unobserved through-plane content under a known sampling mask.
    Problem definition and training objective Eq. (1); if false, stripe completion cannot recover true isotropic anatomy.
  • ad hoc to paper Voxel-wise average of two orthogonal single-view restorations yields a more volumetrically consistent isotropic volume than either view alone for quantitative analysis.
    Inference Eq. (12) and Table 2; simple fusion is a design choice, not a derived optimum.
  • domain assumption Standard image metrics (PSNR/SSIM/GMSD) and Inception-based FID/KID, plus SynthSeg Dice/ASSD/HD99, are valid proxies for restoration quality and downstream utility.
    Sec. 4.4 evaluation protocol; authors note FID/KID are complementary only.
invented entities (1)
  • CRIS stripe-completion formulation (blank-slice insertion operator Ig + validity-masked 2D completion + multi-view fusion)
    purpose: Recast anisotropic 3D isotropic restoration as self-supervised 2D periodic stripe completion without interpolated pseudo-observations.
    Methodological construct introduced in Sec. 3.2–3.6; evaluated empirically, not an independent physical entity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CRIS: Cross-Plane Self-Supervised Isotropic Restoration for Anisotropic Volumetric Imaging Across Modalities." pith.science (2026). https://pith.science/paper/K6DRTNIT

@misc{pith2026260615967,
  author       = {Pith},
  title        = {Pith review of: CRIS: Cross-Plane Self-Supervised Isotropic Restoration for Anisotropic Volumetric Imaging Across Modalities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K6DRTNIT}},
  note         = {Machine review of arXiv:2606.15967}
}
read the original abstract

Anisotropic volumetric acquisitions are common in clinical MRI and volume electron microscopy (vEM), where sparse through-plane sampling creates thick slices or sections that degrade orthogonal reformats and downstream analysis. We present CRIS, a cross-plane self-supervised framework for isotropic restoration without paired isotropic ground truth. CRIS casts 3D restoration as 2D stripe completion on orthogonal reformats of an isotropic grid: high-resolution in-plane slices are synthetically degraded and periodically masked for training, while at inference blank slices define the isotropic grid, two orthogonal reformats are restored, and predictions are fused by multi-view averaging. We evaluate CRIS on two MRI cohorts and two microscopy benchmarks up to 8x anisotropy. On brain MRI, CRIS achieves 32.921 +/- 0.436 dB PSNR and 0.963 +/- 0.003 SSIM, outperforming interpolation, ECLARE, SMORE4, SIMPLE, SA-INR, and ATME, and gives the best segmentation consistency (Dice 0.940 +/- 0.004, ASSD 0.245 +/- 0.014 mm, HD99 1.275 +/- 0.061 mm). On reference-free abdominal MRI, CRIS reduces FID/KID to 48.71/0.023, outperforming interpolation, ECLARE, SMORE4, and SIMPLE. On vEM, CRIS achieves 29.100 dB/0.830 3D PSNR/SSIM at 4x and 26.874 dB/0.722 at 8x on EPFL, and 21.935 +/- 0.437 dB/0.696 +/- 0.024 on noisy hemibrain data. In a dedicated robustness experiment, one variable-gap CRIS model evaluated across gap factors 3-7 and coronal, axial, and sagittal degradations maintained higher PSNR/SSIM than interpolation (36.36-31.14 dB and 0.977-0.932 vs. 33.07-27.85 dB and 0.951-0.853). These results support CRIS as a modality-flexible route to isotropic restoration without paired isotropic targets or configuration-specific retraining. Code is available at https://github.com/adi-hatav/CRIS.

Figures

Figures reproduced from arXiv: 2606.15967 by the authors.

Figure 1
Figure 1. CRIS overview during training and inference. Training: high-resolution in-plane slices are extracted and transformed into self-supervised input–target pairs through modality-specific degradation and periodic masking. Inference: blank slices are inserted to define the isotropic grid, orthogonal reformats are restored independently by the trained model, and the results are fused via multi-view averaging to obtain the … view at source ↗
Figure 2
Figure 2. Validation-set 𝜎 calibration for MRI slice-select simulation. (A) On the brain MRI validation set, GT-referenced FID attains its best value near 𝜎 ≈ 1.67, whereas in-plane-referenced FID misleadingly favors 𝜎 = 0. (B) Brain test-set PSNR also peaks near 𝜎 ≈ 1.67. (C) On the abdominal MRI validation set, in-plane-referenced FID again favors 𝜎 = 0, showing the same bias. These validation trends motivate the physics-in… view at source ↗
Figure 3
Figure 3. Brain MRI qualitative comparison across sagittal and axial planes from an anisotropic coronal source volume with gap factor 𝑔 = 5. Columns are ordered left-to-right as Interpolation, SMORE4, CRIS, and isotropic reference. A. Ahituv et al.: Preprint Page 10 [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Brain MRI robustness across anisotropy severity and degradation orientation. A single CRIS model was trained once using coronal synthetic degradation with randomized training-time gap factors 𝑔 ∈ {3, 4, 5, 6, 7} in the self-supervised degradation process. The trained m…
Figure 5
Figure 5. Figure 5: Abdominal MRI qualitative comparison across sagittal and axial planes from a held-out coronal anisotropic source volume with typical gap factor 𝑔 ≈ 6. Columns are ordered left-to-right as Interpolation, SIMPLE, SMORE4, and CRIS [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison on the EPFL FIB-SEM dataset in the sagittal view under simulated 4× and 8× through-plane anisotropy. Columns are ordered left-to-right as interpolation, NIIV, vEMINR XZ, vEMINR YZ, CRIS, and isotropic ground truth. CRIS preserves ultrastructural …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 21 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , author Royer, L

    author Batson, J. , author Royer, L. , year 2019 . title Noise2self: Blind denoising by self-supervision , in: booktitle Proceedings of the 36th International Conference on Machine Learning (ICML) , pp. pages 524--533

    2019

  3. [3]

    , author Shteynman, Y

    author Benisty, R. , author Shteynman, Y. , author Porat, M. , author Ilivitzki, A. , author Freiman, M. , year 2026 . title Simple: Simultaneous multi-plane self-supervised learning for isotropic mri restoration from anisotropic data , in: booktitle Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025, Proceedings , pp. pages 551--56...

    work page doi:10.1007/978-3-032-05169-1_53 2026

  4. [4]

    , author Greve, D.N

    author Billot, B. , author Greve, D.N. , author Puonti, O. , author Thielscher, A. , author Van Leemput, K. , author Fischl, B. , author Dalca, A.V. , author Iglesias, J.E. , year 2023 a. title Synthseg: Segmentation of brain mri scans of any contrast and resolution without retraining . journal Medical Image Analysis volume 86 , pages 102789 . :10.1016/j....

    work page doi:10.1016/j.media.2023.102789 2023

  5. [5]

    , author Magdamo, C

    author Billot, B. , author Magdamo, C. , author Arnold, S.E. , author Das, S. , author Iglesias, J.E. , year 2023 b. title Robust machine learning segmentation for large-scale analysis of heterogeneous clinical brain mri datasets . journal Proceedings of the National Academy of Sciences volume 120 , pages e2216399120 . :10.1073/pnas.2216399120

    doi:10.1073/pnas.2216399120 2023

  6. [6]

    , author Wang, Y

    author Cao, H. , author Wang, Y. , author Chen, J. , author Jiang, D. , author Zhang, X. , author Tian, Q. , author Wang, M. , year 2021 . title Swin-unet: Unet-like pure transformer for medical image segmentation . journal arXiv preprint arXiv:2105.05537 :10.48550/arXiv.2105.05537

    arXiv doi:10.48550/arxiv.2105.05537 2021

  7. [7]

    , author Liu, S

    author Chen, Y. , author Liu, S. , author Wang, X. , year 2021 . title Learning continuous image representation with local implicit image function , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 8628--8638 . :10.1109/CVPR46437.2021.00852

    doi:10.1109/cvpr46437.2021.00852 2021

  8. [8]

    title Electron microscopy dataset (ca1 hippocampus fib-sem subvolumes)

    author CVLab, EPFL , year 2025 . title Electron microscopy dataset (ca1 hippocampus fib-sem subvolumes) . howpublished https://www.epfl.ch/labs/cvlab/data/data-em/ . note Accessed 2026-04-04

    2025

Show all 51 references
  1. [9]

    , author Fu, X

    author Deng, S. , author Fu, X. , author Xiong, Z. , author Chen, C. , author Liu, D. , author Chen, X. , author Ling, Q. , author Wu, F. , year 2020 . title Isotropic reconstruction of 3d em images with unsupervised degradation learning , in: booktitle Medical Image Computing...

    2020 doi

  2. [10]

    , author O'Connor, D

    author Di Martino, A. , author O'Connor, D. , author Chen, B. , author Alaerts, K. , author Anderson, J.S. , author Assaf, M. , et al., year 2017 . title Enhancing studies of the connectome in autism using the autism brain imaging data exchange ii . journal Scientific Data vol...

    2017 doi

  3. [11]

    , author Yan, C.G

    author Di Martino, A. , author Yan, C.G. , author Li, Q. , author Denio, E. , author Castellanos, F.X. , author Alaerts, K. , author Anderson, J.S. , author Assaf, M. , author Bookheimer, S.Y. , author Dapretto, M. , author Deen, B. , author Delmonte, S. , author Dinstein, I. ...

    2014 doi

  4. [12]

    , author Estroff, J.A

    author Gholipour, A. , author Estroff, J.A. , author Barnewolt, C.E. , author Connolly, S.A. , author Warfield, S.K. , year 2015 . title Super-resolution reconstruction in frequency, image and wavelet domains to reduce through-plane partial voluming in mri , in: booktitle Medi...

    2015 doi

  5. [13]

    , author Estroff, J.A

    author Gholipour, A. , author Estroff, J.A. , author Sahin, M. , author Prabhu, S.P. , author Warfield, S.K. , year 2010 . title Maximum a posteriori estimation of isotropic high-resolution volumetric mri from orthogonal thick-slice scans , in: booktitle Medical Image Computin...

    2010 doi

  6. [14]

    , author Garc \'i a-Ortiz, J.P

    author Gonz \'a lez-Ruiz, V. , author Garc \'i a-Ortiz, J.P. , author Fern \'a ndez-Fern \'a ndez, M.R. , author Fern \'a ndez, J.J. , year 2022 . title Optical flow driven interpolation for isotropic fib-sem reconstructions . journal Computer Methods and Programs in Biomedici...

    2022 doi

  7. [15]

    , author Oz, G

    author Greenspan, H. , author Oz, G. , author Kiryati, N. , author Peled, S. , year 2002 . title Mri inter-slice reconstruction using super-resolution . journal Magnetic Resonance Imaging volume 20 , pages 437--446 . :10.1016/S0730-725X(02)00511-8

    2002 doi

  8. [16]

    , author Bogovic, J.A

    author Heinrich, L. , author Bogovic, J.A. , author Saalfeld, S. , year 2017 . title Deep learning for isotropic super-resolution from non-isotropic 3d electron microscopy , in: booktitle Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2017 , pp. pages 135...

    2017 doi

  9. [17]

    , author Jain, A

    author Ho, J. , author Jain, A. , author Abbeel, P. , year 2020 . title Denoising diffusion probabilistic models . journal arXiv preprint arXiv:2006.11239 :10.48550/arXiv.2006.11239

    2020 doi arXiv

  10. [18]

    , author Billot, B

    author Iglesias, J.E. , author Billot, B. , author Balbastre, Y. , author Magdamo, C. , author Arnold, S.E. , author Das, S. , author Edlow, B.L. , author Alexander, D.C. , author Golland, P. , author Fischl, B. , year 2023 . title Synthsr: A public ai tool to turn heterogeneo...

    2023 doi

  11. [19]

    , author Bernstein, M.A

    author Jack, C.R. , author Bernstein, M.A. , author Fox, N.C. , author Thompson, P. , author Alexander, G. , author Harvey, D. , author Borowski, B. , author Britson, P.J. , author Whitwell, J.L. , author Ward, C. , author Dale, A.M. , author Felmlee, J.P. , author Gunter, J.L...

    2008 doi

  12. [20]

    title Flyem hemibrain project

    author Janelia Research Campus, HHMI , year 2026 . title Flyem hemibrain project . howpublished https://www.janelia.org/project-team/flyem/hemibrain . note Accessed 2026-04-04

    2026

  13. [21]

    , author Dai, B

    author Jiang, L. , author Dai, B. , author Wu, W. , author Loy, C.C. , year 2021 . title Focal frequency loss for image reconstruction and synthesis , in: booktitle Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pp. pages 13899--13909 . :10.11...

    2021 doi

  14. [22]

    , author Vasylechko, S

    author Khateri, M. , author Vasylechko, S. , author Ghahremani, M. , author Timms, L. , author Kocanaogullari, D. , author Warfield, S.K. , author Jaimes, C. , author Karimi, D. , author Sierra, A. , author Tohka, J. , author Kurugol, S. , author Afacan, O. , year 2025 . title...

    2025 doi

  15. [23]

    , author Thanki, A

    author Kim, H. , author Thanki, A. , author van Herten, R. , author Margolis, D. , author Sabuncu, M.R. , year 2026 . title Single-subject multi-view mri super-resolution via implicit neural representations . journal arXiv preprint arXiv:2603.22627

    2026

  16. [24]

    , author Buchholz, T.O

    author Krull, A. , author Buchholz, T.O. , author Jug, F. , year 2019 . title Noise2void -- learning denoising from single noisy images , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 2124--2132 . :10.1109/CV...

    2019 doi

  17. [25]

    , author Lin, Y

    author Liu, Z. , author Lin, Y. , author Cao, Y. , author Hu, H. , author Wei, Y. , author Zhang, Z. , author Lin, S. , author Guo, B. , year 2021 . title Swin transformer: Hierarchical vision transformer using shifted windows , in: booktitle Proceedings of the IEEE/CVF Intern...

    2021

  18. [26]

    , author Wang, J

    author Lu, Z. , author Wang, J. , author Li, Z. , author Ying, S. , author Wang, J. , author Shi, J. , author Shen, D. , year 2023 . title Two-stage self-supervised cycle-consistency transformer network for reducing slice gap in mr images . journal IEEE Journal of Biomedical a...

    2023 doi

  19. [27]

    , author Li, Y

    author Lucchi, A. , author Li, Y. , author Fua, P. , year 2013 . title Learning for structured prediction using approximate subgradient descent with working sets , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 19...

    2013 doi

  20. [28]

    , author Coup \'e , P

    author Manj \'o n, J.V. , author Coup \'e , P. , author Buades, A. , author Fonov, V. , author Collins, D.L. , author Robles, M. , year 2010 . title Non-local mri upsampling . journal Medical Image Analysis volume 14 , pages 784--792 . :10.1016/j.media.2010.05.010

    2010 doi

  21. [29]

    , author Gan, Y

    author Pan, M. , author Gan, Y. , author Zhou, F. , author Liu, J. , author Zhang, Y. , author Wang, A. , author Zhang, S. , author Li, D. , year 2023 . title Diffuseir: Diffusion models for isotropic reconstruction of 3d microscopic images , in: booktitle Medical Image Comput...

    2023 doi

  22. [30]

    , author Collinson, L.M

    author Peddie, C.J. , author Collinson, L.M. , year 2014 . title Exploring the third dimension: Volume electron microscopy comes of age . journal Micron volume 61 , pages 9--19 . :10.1016/j.micron.2014.01.009

    2014 doi

  23. [31]

    , author Genoud, C

    author Peddie, C.J. , author Genoud, C. , author Kreshuk, A. , author Meechan, K. , author Micheva, K.D. , author Narayan, K. , author Pape, C. , author Parton, R.G. , author Schieber, N.L. , author Schwab, Y. , author Titze, B. , author Verkade, P. , author Collinson, L.M. , ...

    2022 doi

  24. [32]

    , author Lin, W.A

    author Peng, C. , author Lin, W.A. , author Liao, H. , author Chellappa, R. , author Zhou, S.K. , year 2020 . title Saint: Spatially aware interpolation network for medical slice synthesis , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Re...

    2020 doi

  25. [33]

    , author Han, S

    author Remedios, S.W. , author Han, S. , author Zuo, L. , author Carass, A. , author Pham, D.L. , author Prince, J.L. , author Dewey, B.E. , year 2023 . title Self-supervised super-resolution for anisotropic mr images with and without slice gap , in: booktitle Simulation and S...

    2023 doi

  26. [34]

    , year 2010

    author Rousseau, F. , year 2010 . title A non-local approach for image super-resolution using intermodality priors . journal Medical Image Analysis volume 14 , pages 594--605 . :10.1016/j.media.2010.04.005

    2010 doi

  27. [35]

    , author Malpica, N

    author Rueda, A. , author Malpica, N. , author Romero, E. , year 2013 . title Single-image super-resolution of brain mr images using overcomplete dictionaries . journal Medical Image Analysis volume 17 , pages 113--132 . :10.1016/j.media.2012.09.003

    2013 doi

  28. [36]

    , author Xu, C.S

    author Scheffer, L.K. , author Xu, C.S. , author Januszewski, M. , author Lu, Z. , author Takemura, S.Y. , author Hayworth, K.J. , et al., year 2020 . title A connectome and analysis of the adult drosophila central brain . journal eLife volume 9 , pages e57443 . :10.7554/eLife.57443

    2020 doi

  29. [37]

    , author Januszewski, M

    author Shapson-Coe, A. , author Januszewski, M. , author Berger, D.R. , author Pope, A. , author Wu, Y. , author Blakely, T. , et al., year 2024 . title A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution . journal Science volume 384 , pages ead...

    2024 doi

  30. [38]

    , author Cohen, N

    author Shocher, A. , author Cohen, N. , author Irani, M. , year 2018 . title ``zero-shot'' super-resolution using deep internal learning , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 3118--3126 . :10.1109/CVPR....

    2018 doi

  31. [39]

    , author Martel, J.N.P

    author Sitzmann, V. , author Martel, J.N.P. , author Bergman, A.W. , author Lindell, D.B. , author Wetzstein, G. , year 2020 . title Implicit neural representations with periodic activation functions , in: booktitle Advances in Neural Information Processing Systems , pp. pages...

    2020

  32. [40]

    , author Rodriguez, A.B

    author Solano-Carrillo, E. , author Rodriguez, A.B. , author Carrillo-Perez, B. , author Steiniger, Y. , author Stoppe, J. , year 2023 . title Look atme: The discriminator mean entropy needs attention , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision an...

    2023

  33. [41]

    , author Mao, X

    author Song, H. , author Mao, X. , author Yu, J. , author Li, Q. , author Wang, Y. , year 2024 . title I ^ 3 net: Inter-intra-slice interpolation network for medical slice synthesis . journal IEEE Transactions on Medical Imaging volume 43 , pages 3306--3318 . :10.1109/TMI.2024.3394033

    2024 doi

  34. [42]

    , author Sohl-Dickstein, J

    author Song, Y. , author Sohl-Dickstein, J. , author Kingma, D.P. , author Kumar, A. , author Ermon, S. , author Poole, B. , year 2020 . title Score-based generative modeling through stochastic differential equations . journal arXiv preprint arXiv:2011.13456 :10.48550/arXiv.2011.13456

    2020 doi arXiv

  35. [43]

    , author Vanhoucke, V

    author Szegedy, C. , author Vanhoucke, V. , author Ioffe, S. , author Shlens, J. , author Wojna, Z. , year 2016 . title Rethinking the inception architecture for computer vision , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR...

    2016 doi

  36. [44]

    , author Liang, Y

    author Troidl, J. , author Liang, Y. , author Beyer, J. , author Tavakoli, M. , author Danzl, J.G. , author Hadwiger, M. , author Pfister, H. , author Tompkin, J. , year 2026 a. title niiv: Interactive self-supervised neural implicit isotropic volume reconstruction , in: bookt...

    2026 doi

  37. [45]

    , author Liang, Y

    author Troidl, J. , author Liang, Y. , author Beyer, J. , author Tavakoli, M. , author Danzl, J.G. , author Hadwiger, M. , author Pfister, H. , author Tompkin, J. , year 2026 b. title niiv-miccai: Interactive self-supervised neural implicit isotropic volume reconstruction . ho...

    2026

  38. [46]

    , author Vedaldi, A

    author Ulyanov, D. , author Vedaldi, A. , author Lempitsky, V. , year 2018 . title Deep image prior , in: booktitle Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 9446--9454 . :10.1109/CVPR.2018.00984

    2018 doi

  39. [47]

    , author Wang, S

    author Wang, X. , author Wang, S. , author Xiong, H. , author Xuan, K. , author Zhuang, Z. , author Liu, M. , author Shen, Z. , author Zhao, X. , author Zhang, L. , author Wang, Q. , year 2024 . title Spatial attention-based implicit neural representation for arbitrary reducti...

    2024 doi

  40. [48]

    , author Bovik, A.C

    author Wang, Z. , author Bovik, A.C. , author Sheikh, H.R. , author Simoncelli, E.P. , year 2004 . title Image quality assessment: From error visibility to structural similarity . journal IEEE Transactions on Image Processing volume 13 , pages 600--612 . :10.1109/TIP.2003.819861

    2004 doi

  41. [49]

    , author Zhang, L

    author Xue, W. , author Zhang, L. , author Mou, X. , author Bovik, A.C. , year 2014 . title Gradient magnitude similarity deviation: A highly efficient perceptual image quality index . journal IEEE Transactions on Image Processing volume 23 , pages 684--695 . :10.1109/TIP.2013.2293423

    2014 doi

  42. [50]

    , author Huo, J

    author Yang, J. , author Huo, J. , author Liu, M. , author Feng, C. , author Zhang, Y. , author Pan, G. , author Meng, W. , author Han, R. , year 2026 . title veminr: Ultra-fast isotropic reconstruction for volume electron microscopy with implicit neural representation . journ...

    2026 doi

  43. [51]

    , author Dewey, B.E

    author Zhao, C. , author Dewey, B.E. , author Pham, D.L. , author Calabresi, P.A. , author Reich, D.S. , author Prince, J.L. , year 2021 . title Smore: A self-supervised anti-aliasing and super-resolution algorithm for mri using deep learning . journal IEEE Transactions on Med...

    2021 doi

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.