Pith. sign in

REVIEW 3 major objections 5 minor 27 references

Do 3D Medical Foundation Models See Through MRI Artifacts? A Controlled Study of Representation Robustness

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read MRI artifacts disturb 3D medical encoders in model- and artifact-specific ways, usually by geometrically drifting the feature space rather than collapsing its dimensionality.

desk verdict A useful first controlled benchmark of 3D medical encoder robustness to MRI artifacts, with honest caveats but unresolved preprocessing confounds and no uncertainty quantification. read the letter →

arxiv 2608.06613 v1 pith:TISNWF2T submitted 2026-08-06 cs.CV cs.AI

Julia Anna Mielcarz , Daniel Klaaby , Mostafa Mehdipour Ghazi This is my paper · ORCID
classification cs.CVcs.AI
keywords MRIartifactsrepresentationrobustness3Dmedicalfoundationmodelsself-supervisedlearningcenteredkernelalignmentRankMeartifactsimulationBraTS-Africa
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

Pretrained 3D medical encoders are increasingly used as general-purpose feature extractors, and this paper asks whether their internal representations stay stable when MRI scans are corrupted by realistic artifacts. By perturbing BraTS-Africa volumes with seven frequency- and image-domain artifacts at five severity levels and comparing clean versus corrupted embeddings with linear CKA and RankMe, the authors find that robustness is strongly model- and artifact-dependent: 3DINO is the most consistently stable, BrainIAC is highly sensitive, and the other encoders sit in between with distinct profiles. They also find that CKA often drops substantially while RankMe stays flat, so artifacts reorganize representation geometry without collapsing the effective rank. A separate segmentation-consistency check shows that task-level degradation only partially matches representation-level sensitivity. The practical point is that pretraining scale and domain specificity are not enough to guarantee artifact invariance, so explicit robustness evaluation belongs in model selection for heterogeneous MRI settings.

What carries the argument

Controlled artifact simulation is the load-bearing mechanism: a clean volume $I$ is passed through an artifact operator $T_a(I;\theta_{a,\ell},\xi_a)$ implemented either in the frequency domain (k-space spikes, periodic ghosting, Gibbs truncation) or in the image domain (Rician noise, Gaussian blur, bias field, gamma contrast), and only then through each encoder's official preprocessing. Robustness is read off two complementary statistics on matched clean and corrupted embedding matrices. Linear CKA, $\mathrm{CKA}(X,Y) = \| X^\top Y \|_F^2 / (\|X^\top X\|_F \|Y^\top Y\|_F)$ on centered embeddings, measures whether the pairwise geometry of the clean set survives the corruption; RankMe, computed from normalized singular values as $\exp(-\sum_j p_j \log p_j)$ with $p_j = \sigma_j/\sum_k \sigma_k + \varepsilon$, measures whether the effective number of occupied directions survives. Interpreting the two together is what separates geometric drift from dimensional collapse.

What would settle it

Run the same CKA and RankMe benchmark with a single shared preprocessing step for all five encoders; if the model ranking changes, then preprocessing, not the learned representations, produced the reported robustness ordering.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that artifact robustness in 3D medical encoders is neither guaranteed by scale nor by domain specialization: across 95 BraTS-Africa cases, four MRI sequences, seven artifact types, and five severity settings, the five frozen encoders show strongly model- and artifact-dependent behavior. 3DINO, pretrained on roughly 100,000 multi-organ volumes, is the most consistently stable, while BrainIAC, a brain-MRI-specific contrastive ViT, is the most sensitive, with CKA approaching zero under frequency-domain corruptions. The most informative pattern is a dissociation between the two metrics: linear CKA drops substantially for many artifact–model combinations while RankMe stays nearly flat, which the authors read as geometric drift in the embedding space without dimensional collapse. A separate segmentation-consistency analysis with TumorSynth shows Dice declining most under ghosting and Rician noise, but the task-level ordering matches representation-level ordering only partially. The paper's conclusion is therefore that explicit representation-robustness evaluation is needed before deploying these models in heterogeneous MRI settings.

Load-bearing premise

The evaluation assumes that the way each model rescales, crops, or normalizes the image before analysis does not decide the results, even though the corruptions are added before that step.

Editorial extensions

If this is right

  • A model trained on broad multi-organ medical data can be robust to MRI artifacts even when a brain-specific model trained for the same modality is not.
  • Deploying these encoders in heterogeneous clinical settings should include an explicit artifact-robustness check, since neither pretraining scale nor domain specificity predicts stability.
  • Reporting only a spectral metric like RankMe can miss large geometric changes measured by CKA, so robust-representation evaluation should report both.
  • Representation-level sensitivity and task-level segmentation consistency do not move together, so downstream validation cannot substitute for representation analysis.

Reading between the lines

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

  • A direct implication the authors do not pursue is that the CKA-versus-RankMe dissociation predicts stronger degradation for tasks relying on local neighborhood structure, such as retrieval or few-shot classification, than for tasks using global subspace information; a direct test would run the same corruptions through linear-probe and retrieval benchmarks.
  • Because architecture, objective, pretraining domain, and scale differ simultaneously across the five models, the paper cannot identify which design choice drives 3DINO's stability; a controlled ablation varying only the pretraining domain would isolate that factor.
  • Because corruptions pass through each model's official preprocessing, the reported numbers describe end-to-end pipeline robustness rather than pure encoder robustness; a shared-preprocessing variant would split the two contributions.
  • If the pattern generalizes, a hospital using a single foundation model across scanners should expect artifact-induced feature-space shifts even when anatomy is unchanged, making quality-based sample exclusion or artifact-robust fine-tuning a practical next step.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper presents an empirical benchmark of five frozen 3D medical image encoders (3DINO, BrainIAC, BrainFM, NeuroVFM, Neuro-SimCLR) under seven simulated MRI artifacts at five severity levels, using 95 BraTS-Africa cases across four modalities. Robustness is measured by linear CKA between clean and corrupted embedding matrices, RankMe spectral effective rank, UMAP visualization, and the Dice consistency of an independent segmentation model (TumorSynth). The authors report that robustness varies strongly by model and artifact, with 3DINO as the most consistently stable encoder, BrainIAC as the most sensitive, and CKA declining while RankMe remains comparatively stable, which they interpret as geometric drift without dimensional collapse.

Significance. The benchmark addresses a real and under-explored gap: how pretrained 3D medical foundation models behave under clinically relevant MRI artifacts at the representation level rather than through a single downstream task. The design is largely non-circular: artifact parameters are fixed in Table 2, no constants are fitted to the reported metrics, and the model ranking is a falsifiable empirical claim. The inclusion of an independent segmentation-consistency reference and four MRI modalities strengthens the study. However, the headline conclusions about model ordering are, as analyzed below, threatened by a preprocessing confound and by the absence of uncertainty quantification. If these issues are addressed, the paper would be a useful reference for robustness evaluation and model selection in heterogeneous MRI settings.

major comments (3)
  1. [Sections 3.1–3.2 and Appendix A] The central model ranking may be driven by differences in model-specific preprocessing rather than by the learned representations. In Eq. (1) and Eq. (2), artifacts are applied to the common normalized volume I before the model-specific preprocessing operator P_m: z = g_m(f_m(P_m(I))). The preprocessing operators differ drastically across models: 3DINO resizes to 112^3 (A.1), BrainIAC resizes to 96^3 and z-scores non-zero voxels (A.2), BrainFM skull-strips and resamples to 1-mm isotropic (A.3), NeuroVFM applies its own quantization (A.4), and Neuro-SimCLR crops and z-scores (A.5). In particular, 3DINO's large downsampling acts as a low-pass filter that can attenuate k-space spikes, ghosting, and Gibbs ringing before the encoder sees them, which may explain its high CKA on exactly those frequency-domain artifacts. The statement in Section 3.1 that identical preprocessing is applied to clean and corrupted volumes only controls for within-model consistency; it does not isolate the contribution of P_m from that of f_m. The abstract and conclusion claim that 'larger-scale or domain-specific pretraining alone does not guarantee artifact invariance' is therefore not supported unless the preprocessing confound is addressed. I request a control experiment, for example applying a common preprocessing to all models, corrupting after preprocessing, or quantifying the effective artifact amplitude after each P_m, to show that the observed model ordering survives.
  2. [Equation (15)] The normalized singular values are not correctly normalized. In Eq. (15), p_j is defined as sigma_j / sum_k sigma_k + epsilon, so the p_j sum to 1 + r*epsilon rather than 1, and for zero singular values p_j = epsilon enters the entropy as epsilon log epsilon. The text calls these 'normalized singular values' and treats them as a probability distribution, which is inconsistent with the displayed formula. Since RankMe values and the conclusion that RankMe remains stable across corruptions are based on this formula, the definition should be corrected (for example, p_j = sigma_j / (sum_k sigma_k + r*epsilon), or using log(sigma_j + epsilon) inside the entropy), and the value of epsilon should be stated. Without this fix, the reported RankMe numbers are not reproducible from the paper as written.
  3. [Section 3.3 and Figures 1–2] All quantitative claims are based on single point estimates computed from N = 95 volumes per modality, with no confidence intervals, bootstrap resampling, or significance tests. The differences between CKA curves that are described as 'strongly' model- and artifact-dependent may be within sampling variability for this sample size, and the ordering of models in the severity curves is presented without an uncertainty measure. The segmentation-consistency result in Figure 4 similarly reports mean Dice without error bars or paired tests. The authors should add, at minimum, bootstrap confidence intervals over cases (or equivalently a paired significance test for the main model-ranking comparisons), and ideally report variability over the random artifact parameters xi_a in Eq. (2), which currently appear to be fixed but are not described as such.
minor comments (5)
  1. [Section 3.2] For periodic k-space ghosting, the period R = 3 and axis q = x are fixed, and only alpha_gh is varied; since real ghosting depends on the phase-encoding direction and period, a sentence acknowledging this limitation would be useful, and an additional analysis varying R and q would strengthen the generality of the ghosting results.
  2. [Section 3.3] The UMAP description says 'UMAP is run with the default setting' without reporting n_neighbors, min_dist, metric, or random seed; for reproducibility, these parameters should be listed in the appendix.
  3. [Section 3.3] The sentence after Eq. (15), 'RankMe is defined as', is grammatically incomplete and should be rewritten to describe the formula properly.
  4. [Figure 4 caption] The caption states that Dice is 'averaged across labels 1–18 and cases'; please clarify whether the whole-tumor label 18 is included in this average or treated separately, and whether Dice is computed per label and then averaged or computed on the union of labels.
  5. [References] There are formatting errors in the reference list, for example 'PMlR' in reference [12] and 'PmLR' in reference [17]; these should be corrected to 'PMLR'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the robustness ranking is an empirical benchmark with fixed artifact parameters, and the only author self-citations support input-generation formulas rather than the target model ranking.

full rationale

This paper is an empirical benchmark, not a derivation. The claimed results—that 3DINO is the most consistently stable encoder, that BrainIAC is highly artifact-sensitive, and that CKA and RankMe capture complementary failure modes—are obtained by running frozen public encoders on BraTS-Africa volumes corrupted by fixed artifact transformations. No parameter is fitted to the target CKA, RankMe, or Dice outcomes: the artifact severity levels (Table 2) are fixed before any encoder is evaluated, and the metrics are standard, externally defined quantities. The only author self-citations are references [21] and [22], used to motivate the Gibbs-ringing truncation mask (Eq. 7) and the elliptic bias-field model (Eq. 8). Those formulas are standard Fourier-truncation and multiplicative-bias constructions, and they are input generators rather than results that force the observed model ordering; the same corruptions are applied to every encoder before each model-specific preprocessor. Accordingly, none of the load-bearing conclusions reduces to a fitted parameter renamed as a prediction, to a self-definition, or to a uniqueness theorem imported from the authors' prior work. A separate validity concern—that model-specific preprocessing may attenuate artifacts differently and thereby influence the ranking—is a possible confounding factor in the experimental design, not a circularity; it does not make any claim equivalent to its inputs by construction. The paper is self-contained against external benchmarks, and the honest finding is that no significant circularity is present.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claims rest on the validity of the simulated artifacts as proxies for clinical MRI degradation, the representativeness of the BraTS-Africa cohort, and the assumption that model-specific preprocessing does not determine the ranking. The benchmark contains no fitted free parameters; severity levels are fixed by visual inspection before any encoder is run.

free parameters (4)
  • Artifact severity levels L1-L5 (Table 2) = Rician sigma 0.07-0.15; blur sigma 0.60-1.40; bias lambda 0.60-1.40; gamma 0.40-2.00; spike alpha 0.45-1.05; ghost…
    Chosen through preliminary visual inspection, not calibrated to clinical image-quality grades; all cross-artifact and severity comparisons depend on these choices.
  • Ghosting period R and encoding axis q = R=3, q=x
    Fixed by hand; other values would change ghost severity and could alter the model ranking.
  • Gibbs truncation axis = y-axis
    Fixed by hand; truncation along a different axis could change the ringing pattern and severity.
  • Number of k-space spike coefficients M = 2
    Fixed by hand; severity is varied only through amplitude, not through the number of corrupted coefficients.
assumptions (5)
  • domain assumption The seven simulated transforms adequately proxy the MRI artifacts they are named after.
    The whole benchmark measures responses to synthetic corruptions; if the simulations are not clinically representative, the conclusions about real-world robustness do not transfer.
  • domain assumption Each encoder's official preprocessing pipeline is applied identically to clean and corrupted inputs and does not confound the cross-model ranking.
    Equation (1) and Appendix A define model-specific preprocessing; resizing, skull-stripping, and normalization can remove or amplify artifacts before the encoder, so differences across models may originate in preprocessing.
  • domain assumption Linear CKA and RankMe computed on N=95 matched embeddings capture meaningful representation robustness.
    CKA in high-dimensional embeddings with a small sample may be sensitive to sample size, and no confidence intervals or significance tests are provided.
  • domain assumption TumorSynth predictions provide an independent task-level reference.
    TumorSynth's own preprocessing and failure modes are not characterized in this study; Dice consistency is only prediction stability, not segmentation accuracy.
  • standard math The FFT-based artifact operations and the CKA and RankMe definitions are standard mathematical tools used as specified.
    No new mathematics is introduced; the equations follow prior work cited by the authors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Do 3D Medical Foundation Models See Through MRI Artifacts? A Controlled Study of Representation Robustness." pith.science (2026). https://pith.science/paper/TISNWF2T

@misc{pith2026260806613,
  author       = {Pith},
  title        = {Pith review of: Do 3D Medical Foundation Models See Through MRI Artifacts? A Controlled Study of Representation Robustness},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TISNWF2T}},
  note         = {Machine review of arXiv:2608.06613}
}
read the original abstract

Self-supervised 3D medical foundation models are increasingly used as general-purpose feature extractors, yet their sensitivity to MRI artifacts remains poorly understood. We present a controlled evaluation of representation robustness across five pretrained 3D encoders spanning different architectures, objectives, pretraining domains, and dataset scales. Using BraTS-Africa cases with four MRI sequences, we generate seven frequency- and image-domain artifacts at five predefined corruption settings. Robustness is assessed using linear centered kernel alignment (CKA), RankMe, and UMAP, complemented by an independent segmentation-consistency analysis. We find that robustness is strongly model- and artifact-dependent. 3DINO exhibits the most consistently stable representations, while BrainIAC is highly sensitive to several corruptions; NeuroVFM, BrainFM, and Neuro-SimCLR show intermediate but distinct artifact-specific profiles. Across many conditions, CKA decreases substantially while RankMe remains comparatively stable, indicating that artifacts often distort representation geometry without causing dimensional collapse. Segmentation consistency also degrades under corruption, particularly for ghosting and Rician noise, but aligns only partially with representation-level robustness. These findings show that larger-scale or domain-specific pretraining alone does not guarantee artifact invariance and motivate explicit robustness evaluation before deploying 3D foundation models in heterogeneous MRI settings.

Figures

Figures reproduced from arXiv: 2608.06613 by the authors.

Figure 1
Figure 1. Representation robustness to image-domain artifacts on T1 MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Representation robustness to frequency-domain artifacts on T1 MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder. 4.2 Geometric Drift Without Consistent Dimensional Collapse CKA and RankMe reveal complementary aspects of representation robustness. Across many artifact–model combinations, CKA decreases substantially with corruption strength while RankMe remains nearly constant. This… view at source ↗
Figure 3
Figure 3. UMAP projections of T1 representations from five encoders under seven arti￾facts. Clean and artifact-corrupted samples are embedded jointly for each encoder. The T1 UMAP projections in [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Segmentation consistency under artifact corruption. Dice similarity between clean and corrupted TumorSynth predictions is averaged across labels 1–18 and cases. predicted regions outside the main anatomy. The remaining corruptions primar￾ily induce smaller boundary and…
Figure 5
Figure 5. Figure 5: Representation robustness to image-domain artifacts on T1ce MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: Representation robustness to frequency-domain artifacts on T1ce MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Representation robustness to image-domain artifacts on T2 MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Representation robustness to frequency-domain artifacts on T2 MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Representation robustness to image-domain artifacts on FLAIR MRI. Linear CKA and RankMe are shown across five corruption settings for each encoder [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Representation robustness to frequency-domain artifacts on FLAIR MRI. Lin￾ear CKA and RankMe are shown across five corruption settings for each encoder. BrainIAC BrainFM NeuroSimCLR NeuroVFM 3DINO Rician noise Gaussian blur Bias field Gamma contrast K-space spike Moti…
Figure 11
Figure 11. Figure 11: UMAP projections of T1ce representations from five encoders under seven artifacts. Clean and artifact-corrupted samples are embedded jointly for each encoder [PITH_FULL_IMAGE:figures/full_fig_p025_11.png]
Figure 12
Figure 12. Figure 12: UMAP projections of T2 representations from five encoders under seven ar￾tifacts. Clean and artifact-corrupted samples are embedded jointly for each encoder. BrainIAC BrainFM NeuroSimCLR NeuroVFM 3DINO Rician noise Gaussian blur Bias field Gamma contrast K-space spike…
Figure 13
Figure 13. Figure 13: UMAP projections of FLAIR representations from five encoders under seven artifacts. Clean and artifact-corrupted samples are embedded jointly for each encoder [PITH_FULL_IMAGE:figures/full_fig_p026_13.png]
Figure 14
Figure 14. Figure 14: Middle slices of representative T1 MRI under the seven simulated artifacts. Each row corresponds to one artifact type, and columns show the clean image followed by the five predefined corruption settings [PITH_FULL_IMAGE:figures/full_fig_p027_14.png]
Figure 15
Figure 15. Figure 15: Middle slices of representative TumorSynth predictions under artifact corrup￾tion. Each row corresponds to one artifact type, and columns show the 18 predicted anatomical and tumor classes for the clean image and five corruption settings [PITH_FULL_IMAGE:figures/full…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 25 canonical work pages

  1. [1]

    npj Digital Medicine8(1) (2025) 639 1, 4, 5

    Xu, T., Hosseini, S., Anderson, C., Rinaldi, A., Krishnan, R.G., Martel, A.L., Goubran, M.: A generalizable 3D framework and model for self-supervised learning in medical imaging. npj Digital Medicine8(1) (2025) 639 1, 4, 5

    work page 2025

  2. [2]

    Nature Neuroscience (2026) 1–12 1, 4, 5

    Tak, D., Garomsa, B.A., Zapaishchykova, A., Chaunzwa, T.L., Climent Pardo, J.C., Ye, Z., Zielke, J., Ravipati, Y., Pai, S., Vajapeyam, S., et al.: A generalizable foundation model for analysis of human brain MRI. Nature Neuroscience (2026) 1–12 1, 4, 5

    work page 2026

  3. [3]

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

    Kaczmarek, E., Szeto, J., Nichyporuk, B., Arbel, T.: Building a general SimCLR self-supervised foundation model across neurological diseases to advance 3D brain MRI diagnoses. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. (2025) 1310–1319 1, 4, 5

    work page 2025

  4. [4]

    arXiv preprint arXiv:2509.00549 (2025) 1, 4, 5

    Liu, P., Puonti, O., Hu, X., Gopinath, K., Sorby-Adams, A., Alexander, D.C., Kimberly, W.T., Iglesias, J.E.: A modality-agnostic multi-task foundation model for human brain imaging. arXiv preprint arXiv:2509.00549 (2025) 1, 4, 5

    arXiv 2025

  5. [5]

    Nature Medicine (2026) 1–7 1, 4, 5

    Kondepudi, A., Rao, A., Zhao, C., Lyu, Y., Harake, S., Banerjee, S., Ogle, J., Joshi, R., Meissner, A.K., Hou, X., et al.: Health system learning enables generalist neuroimaging models. Nature Medicine (2026) 1–7 1, 4, 5

    work page 2026

  6. [6]

    Polish Journal of Radiology80(2015) 93 2, 5, 6

    Krupa, K., Bekiesińska-Figatowska, M.: Artifacts in magnetic resonance imaging. Polish Journal of Radiology80(2015) 93 2, 5, 6

    work page 2015

  7. [7]

    Journal of the American College of Radiology12(7) (2015) 689–695 2

    Andre, J.B., Bresnahan, B.W., Mossa-Basha, M., Hoff, M.N., Smith, C.P., Anzai, Y., Cohen, W.A.: Toward quantifying the prevalence, severity, and cost associated with patient motion during clinical MR examinations. Journal of the American College of Radiology12(7) (2015) 689–695 2

    work page 2015

  8. [8]

    Nature Machine Intelligence2(11) (2020) 665–673 2

    Geirhos, R., Jacobsen, J.H., Michaelis, C., Zemel, R., Brendel, W., Bethge, M., Wichmann, F.A.: Shortcut learning in deep neural networks. Nature Machine Intelligence2(11) (2020) 665–673 2

    work page 2020

Show all 27 references
  1. [9]

    arXiv preprint arXiv:1910.04597 (2019) 2

    Glocker, B., Robinson, R., Castro, D.C., Dou, Q., Konukoglu, E.: Machine learning with multi-site imaging data: An empirical study on the impact of scanner effects. arXiv preprint arXiv:1910.04597 (2019) 2

    2019 arXiv

  2. [10]

    Frontiers in Systems Neuro- science2(2008) 249 2

    Kriegeskorte, N., Mur, M., Bandettini, P.A.: Representational similarity analysis- connecting the branches of systems neuroscience. Frontiers in Systems Neuro- science2(2008) 249 2

    2008

  3. [11]

    Advances in Neural Information Processing Systems30(2017) 2

    Raghu, M., Gilmer, J., Yosinski, J., Sohl-Dickstein, J.: SVCCA: Singular vec- tor canonical correlation analysis for deep learning dynamics and interpretability. Advances in Neural Information Processing Systems30(2017) 2

    2017

  4. [12]

    In: International Conference on Machine Learning, PMlR (2019) 3519–3529 2, 9

    Kornblith, S., Norouzi, M., Lee, H., Hinton, G.: Similarity of neural network representations revisited. In: International Conference on Machine Learning, PMlR (2019) 3519–3529 2, 9

    2019

  5. [13]

    In: International Conference on Machine Learning, PMLR (2023) 10929–10974 2, 9

    Garrido, Q., Balestriero, R., Najman, L., Lecun, Y.: RankMe: Assessing the down- stream performance of pretrained self-supervised representations by their rank. In: International Conference on Machine Learning, PMLR (2023) 10929–10974 2, 9

    2023

  6. [14]

    Journal of Open Source Software3(29) (2018) 861 2, 10

    McInnes, L., Healy, J., Saul, N., Großberger, L.: UMAP: Uniform manifold ap- proximation and projection. Journal of Open Source Software3(29) (2018) 861 2, 10

    2018

  7. [15]

    Multivariate Behavioral Research56(3) (2021) 527–542 2

    Del Giudice, M.: Effective dimensionality: A tutorial. Multivariate Behavioral Research56(3) (2021) 527–542 2

    2021

  8. [16]

    Transactions on Machine Learning Research Journal (2024) 4

    Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al.: DINOv2: Learning 3D Foundation Models Under MRI Artifacts 17 robust visual features without supervision. Transactions on Machine Learning Re...

    2024

  9. [17]

    In: International Conference on Machine Learning, PmLR (2020) 1597–1607 4

    Chen, T., Kornblith, S., Norouzi, M., Hinton, G.: A simple framework for con- trastive learning of visual representations. In: International Conference on Machine Learning, PmLR (2020) 1597–1607 4

    2020

  10. [18]

    Medical Physics12(2) (1985) 143–151 6

    Wood, M.L., Henkelman, R.M.: MR image artifacts from periodic motion. Medical Physics12(2) (1985) 143–151 6

    1985

  11. [19]

    International Journal of Biomedical Imaging2008(1) (2008) 184123 6

    Block, K.T., Uecker, M., Frahm, J.: Suppression of MRI truncation artifacts using total variation constrained data extrapolation. International Journal of Biomedical Imaging2008(1) (2008) 184123 6

    2008

  12. [20]

    Magnetic Resonance in Medicine76(5) (2016) 1574– 1581 6

    Kellner, E., Dhital, B., Kiselev, V.G., Reisert, M.: Gibbs-ringing artifact removal based on local subvoxel-shifts. Magnetic Resonance in Medicine76(5) (2016) 1574– 1581 6

    2016

  13. [21]

    In: Scandinavian Conference on Image Analysis, Springer (2025) 161–176 6, 7

    Mehdipour Ghazi, M., Nielsen, M.: FAST-AID Brain: Fast and accurate seg- mentation tool using artificial intelligence developed for brain. In: Scandinavian Conference on Image Analysis, Springer (2025) 161–176 6, 7

    2025

  14. [22]

    In: MICCAI Workshop on Deep Generative Models, Springer (2024) 191–201 6, 7

    Van Eeden Risager, K., Gholamalizadeh, T., Mehdipour Ghazi, M.: Non-reference quality assessment for medical imaging: Application to synthetic brain MRIs. In: MICCAI Workshop on Deep Generative Models, Springer (2024) 191–201 6, 7

    2024

  15. [23]

    IEEE Transactions on Medical Imaging17(1) (1998) 87–97 6

    Sled, J.G., Zijdenbos, A.P., Evans, A.C.: A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Transactions on Medical Imaging17(1) (1998) 87–97 6

    1998

  16. [24]

    Computer Methods and Programs in Biomedicine208(2021) 106236 6

    Pérez-García, F., Sparks, R., Ourselin, S.: TorchIO: a Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning. Computer Methods and Programs in Biomedicine208(2021) 106236 6

    2021

  17. [25]

    Journal of Magnetic Resonance Imaging32(5) (2010) 1197–1208 7

    Hui, C., Zhou, Y.X., Narayana, P.: Fast algorithm for calculation of inhomogeneity gradient in magnetic resonance imaging data. Journal of Magnetic Resonance Imaging32(5) (2010) 1197–1208 7

    2010

  18. [26]

    Radiology: Imaging Cancer8(2) (2026) e250222 10

    Wu, J., Billot, B., Zhao, F., Li, J., Liu, Y., Zhuo, Z., Pontillo, G., Liu, X., Lin, S., Zhang, X., et al.: TumorSynth: Integrated brain tumor and tissue segmentation on brain MRI scans of any resolution and contrast. Radiology: Imaging Cancer8(2) (2026) e250222 10

    2026

  19. [27]

    Adewole, M., Rudie, J.D., Gbadamosi, A., Zhang, D., Raymond, C., Ajigboto- shso, J., Toyobo, O., Aguh, K., Omidiji, O., Akinola, R., et al.: The BraTS-Africa dataset: expanding the brain tumor segmentation data to capture African popula- tions. Radiology: Artificial Intelligen...

    2025

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.