Mind the Detail: Uncovering Clinically Relevant Image Details in Accelerated MRI with Semantically Diverse Reconstructions

Christian F. Baumgartner; Christian Schlarmann; Jan Nikolas Morshuis; Matthias Hein; Thomas K\"ustner

REVIEW 4 major objections 5 minor 2 cited by

Mind the Detail: Uncovering Clinically Relevant Image Details in Accelerated MRI with Semantically Diverse Reconstructions

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Producing multiple semantically different, data-consistent MRI reconstructions reveals pathologies that a single accelerated reconstruction misses.

desk verdict Smart idea for semantically diverse MRI reconstructions, but the detector-based evaluation shares labels with the objective and lacks statistical support, so the clinical claim is overreaching. read the letter →

arxiv 2507.00670 v1 pith:452MGXKL submitted 2025-07-01 eess.IV cs.CV

Jan Nikolas Morshuis , Christian Schlarmann , Thomas Küstner , Christian F. Baumgartner , Matthias Hein This is my paper · ORCID

classification eess.IVcs.CV

keywords acceleratedMRIreconstructionpathologydetectionfalsenegativessemanticdiversitydataconsistencyadversarialrobustnessdiffusionmodels

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

The paper's goal is to show that accelerated MRI reconstruction, even with modern diffusion-based methods, can silently drop small or rare pathologies, and that this risk is reduced by examining a set of reconstructions instead of a single one. It introduces Semantically Diverse Reconstructions (SDR), a post-hoc procedure that takes any initial reconstruction and produces additional reconstructions which are forced to be semantically different in candidate pathology regions while remaining fully consistent with the measured k-space data. On the fastMRI+ knee dataset, using a separately trained Faster-RCNN detector, SDR reconstruction sets detect more true pathologies (higher recall) than the original reconstructions, repeated diffusion sampling, or E2E-VarNet ensembles, and mean average precision does not drop but improves. If the result holds, SDR offers a practical route to surfacing diagnostic uncertainty in accelerated MRI without changing the acquisition.

What carries the argument

The central mechanism is an iterative projected-gradient loop that maximizes the semantic distance $d(\hat{x}^{(i)}, \hat{x}^{(j)}) = \sum_{b \in \mathcal{B}} \| f_\phi(\hat{x}^{(i)}, b) - f_\phi(\hat{x}^{(j)}, b) \|_2$ over a set of proposal boxes $\mathcal{B}$, where $f_\phi$ is a box-feature encoder built on a ViTDet detector with an adversarially robust ViT backbone. Each update is followed by a projection $P_{\mathcal{B}(\hat{x}^{(1)}, r)}$ onto the $\ell^2$-ball of radius $r$ around the initial reconstruction and a data-consistency projection that substitutes measured k-space values; the robust backbone is what turns the feature-space gradient into a semantically meaningful image change. Proposal boxes come either from the ViTDet region proposal network (SDR-A) or from a manual annotation (SDR-M).

What would settle it

Show the fastMRI+ test slices where SDR recovers a pathology that the baseline reconstruction missed to a panel of radiologists, without telling them which reconstruction came from SDR, and ask them to mark the pathologies they can see; if the radiologists do not see the additional pathologies, the claim that SDR uncovers clinically relevant image details is falsified.

Watch

Extended reading notes

Core claim

The paper demonstrates a method rather than a new theory: given an initial reconstruction $\hat{x}^{(1)}$ and a set of candidate bounding boxes in likely pathology regions, SDR generates $N_\mathrm{rec}-1$ additional reconstructions by optimizing each new image so that the $\ell^2$ distance between the ViTDet box features of that image and the features of all previously generated reconstructions is maximized. A projection onto the $\ell^2$-ball of radius $r$ around $\hat{x}^{(1)}$ and a data-consistency step that re-inserts measured k-space values keep every reconstruction both close to the original and fully consistent with the measurement. The paper reports that an adversarially robust ViT backbone is required for these feature gradients to be semantically meaningful, and that the resulting reconstruction set achieves higher pathology recall than repeated sampling of the diffusion-based DDIP and CSGM baselines or retrained E2E-VarNet ensembles, while also improving mean average precision.

Load-bearing premise

The evaluation rests on the premise that a Faster-RCNN detector trained on fastMRI+ pathology annotations is a faithful proxy for clinical diagnosis; if detector recall does not track radiologist-visible pathologies, the conclusion that SDR reduces false-negative diagnoses is unsupported even if the measured recall improvement is real.

Editorial extensions

If this is right

Applying SDR to any base reconstruction method (diffusion-based or deterministic) raises the pathology recall of the reconstruction set above the single reconstruction and above repeated sampling of the same method.
SDR does not buy recall by adding false positives: on fastMRI+ it also improves mean average precision at IoU 0.25.
Because SDR runs in about three seconds per image, it is fast enough to be used post-hoc in a clinical workflow, unlike diffusion sampling which can take minutes.
The automatic mode (SDR-A) works without any user input, while the manual mode (SDR-M) lets a clinician probe a specific region of concern.
Adversarial robustness of the feature backbone is not optional: using the original non-robust ViTDet does not produce recall or mAP gains over repeated sampling.

Reading between the lines

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

The same recipe -- maximize distance in a robust feature space under data-consistency constraints -- could transfer to other inverse problems such as CT reconstruction or super-resolution, provided a robust encoder exists for that modality.
A natural next test is to swap the generation encoder (ViTDet) for another robust feature space (for example, a robust image encoder not trained for detection) to see whether the recall gain is tied to detection-specific features or to semantic alignment in general.
The paper measures recall on the same fastMRI+ annotation distribution used to train the evaluation detector, so the sharpest additional check would be on rare or unseen pathology classes, where the 'small and rare pathologies' motivation is strongest.

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.

Desk Editor's Note

read the letter

Hi [Colleague],

You should know about this one: it proposes a genuinely new post-hoc method, SDR, that takes any accelerated MRI reconstruction and generates several additional reconstructions that are all data-consistent but semantically different, specifically in regions where pathologies might lurk. The trick is to maximize feature distance in a robust ViTDet backbone, and the authors show this can uncover pathologies missed by the original reconstruction or by repeated sampling from diffusion models. The idea is clever, the method is clearly described, they ship code, and it runs about 3 seconds per image, which makes it practical.

What’s actually new: previous work either didn’t generate multiple reconstructions or, like diffusion models, sampled from the posterior but rarely hit the rare pathologies. SDR deliberately pushes reconstructions in semantically distinct directions, and the use of an adversarially robust detector for semantic distance is a sensible move. The manual and automatic proposal modes are a nice touch.

The soft spots are all in the evaluation. The main evidence is detector recall and mAP from a Faster-RCNN trained on fastMRI+ pathology labels, but the SDR objective itself uses a ViTDet fine-tuned on those same labels. So the generator and evaluator share a training distribution, and the measured recall gains could partly reflect optimizing toward that label distribution rather than toward anatomy a radiologist would flag. The paper calls the evaluation "independent" because a different architecture is used, but that doesn’t establish causal independence. This isn’t a fatal flaw—it’s a serious confound that needs addressing with either a reader study, an independent annotation source, or at least a demonstration that random data-consistent perturbations don’t achieve similar recall. That control is missing, which is a shame because it would isolate the contribution of the semantic objective.

Also, no error bars, confidence intervals, or significance tests appear anywhere, yet the abstract says "significantly reduces." That’s unsupported. The qualitative examples are suggestive but not convincing on their own.

The comparison baselines are okay but not exhaustive; they don’t compare to other diversity-focused methods, like the one from Cohen et al. [5] they cite.

So: the core algorithmic idea is interesting and likely useful for uncertainty visualization, but the clinical claim—reducing false-negative diagnoses—exceeds the evidence. The paper deserves a careful peer review, but it needs major experimental revision before I’d trust the headline result. If you’re working on MRI reconstruction or uncertainty quantification, it’s worth reading for the method; don’t cite it yet as evidence of clinical benefit.

Referee Report

4 major / 5 minor

Summary. The paper addresses the clinical risk that accelerated MRI reconstruction may miss small or rare pathologies. It proposes Semantically Diverse Reconstructions (SDR), a post-hoc method that takes an initial reconstruction and generates additional reconstructions that are data-consistent but semantically different in regions proposed by an object detector or by a user. Semantic diversity is measured as feature distance in a ViTDet backbone that is adversarially fine-tuned for robust gradients. The method is evaluated on fastMRI+ knee data with three base reconstruction techniques (DDIP, CSGM, E2E-VarNet), using a separately trained Faster-RCNN detector as a proxy for clinical reading. The reported results show higher recall and comparable or better mAP0.25 for SDR compared with single reconstructions and with repeated sampling from diffusion models.

Significance. The problem addressed is important: reconstruction quality metrics do not capture whether clinically relevant information is preserved, and false-negative diagnoses are a concrete risk in accelerated MRI. The proposed method is interesting because it is post-hoc, applicable to any reconstruction method, and relatively fast, and the authors provide code. The paper also makes a useful conceptual contribution by shifting evaluation from image quality toward pathology detection. However, the current evidence is not yet sufficient to support the strong clinical claims in the abstract. The main quantitative comparison rests on a single bar chart without error bars or significance tests, the generative objective and the evaluation detector are trained on the same fastMRI+ label distribution, and there is no ablation against random data-consistent perturbations. If these gaps are addressed, the contribution could be valuable to the medical imaging community.

major comments (4)

[Section 4, Fig. 4] The central quantitative claim that SDR 'significantly reduces' false negatives is not supported by the reported statistics. Fig. 4 shows single bars without error bars, confidence intervals, or significance tests. Please report mean and variance across subjects or repeated runs and provide a paired statistical test (e.g., Wilcoxon signed-rank) for recall and mAP0.25 between SDR and each baseline.
[Sections 3.3, 3.4, and 4] There is a circularity concern between the generative objective and the evaluation metric. The semantic distance in Eq. (3) is computed in the feature space of a ViTDet fine-tuned on fastMRI+, and the success metric is the recall of a Faster-RCNN trained on the same fastMRI+ annotations. Even though the two detectors are architecturally different, they share the same label distribution and are trained on the same pathology annotations. Thus the reported recall gains may reflect optimization against the evaluation signal rather than recovery of pathology that a clinician would confirm. Please provide an evaluation with an independent annotation source (e.g., a small reader study or a held-out annotation set) or otherwise demonstrate that the gains persist when the evaluation signal is not the training signal of the feature encoder.
[Section 3.2 and Fig. 4] No ablation against random data-consistent perturbations is reported. Because the method projects onto an ℓ2 ball of radius r=3 around the initial reconstruction and then applies data consistency, a necessary control is to generate the same number of reconstructions with random perturbations inside the same ball, followed by the same data-consistency projection, and to measure recall and mAP. Without this control, the specific contribution of the semantic diversity objective in Eq. (2) cannot be isolated from the effect of merely exploring a set of data-consistent solutions.
[Section 4] The evaluation equates object-detector recall with a reduction in false-negative diagnoses. The paper itself states that clinical assessment would ideally require clinician input, but no clinical validation is provided. The abstract's phrasing about 'false-negative diagnoses' exceeds what the current evidence supports. Please either temper the clinical claims or add evidence that the detector's detections correspond to radiologist-identified pathologies, for example by reporting per-class recall and a qualitative evaluation on a small clinician-annotated subset.

minor comments (5)

[Algorithm 1 and Section 3.2] The variable Nopt appears in Algorithm 1 but is not defined in the text; please define it and clarify the relationship between Eq. (2) and the initialization of each reconstruction with Gaussian noise.
[Throughout] The name of the baseline is written inconsistently as 'E2E-VarNet' and 'E2E-Varnet'; please unify the spelling.
[Fig. 4] The bar chart does not show numerical values, which makes it difficult for readers to assess the magnitude of the reported differences; consider adding value labels or a corresponding table.
[Fig. 4 caption] The caption states that SDR 'generates clinically relevant reconstructions outperforming resampling', which is stronger than the evidence presented; please soften the wording to match the statistical support.
[References] Reference [5] is cited as evidence that diffusion models rarely sample less likely but diagnostically relevant solutions; a sentence elaborating on this finding would help the reader understand the motivation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SDR objective and the evaluation detector are distinct, with held-out subject-wise splits; shared fastMRI+ annotations are an evaluation-proxy concern, not a definitional reduction.

full rationale

The derivation chain is self-contained. SDR's objective (Eq. 3) maximizes pairwise feature distance in a ViTDet box-feature encoder, which is a diversity objective rather than the evaluation metric (Faster-RCNN recall and mAP0.25). The evaluation detector is described as a separate network trained for objective evaluation independent of the ViTDet used for generation (Section 4), and the data split is subject-wise into test and validation sets. The robust backbone is trained on ImageNet data (Section 3.3), which does not use the fastMRI+ pathology labels. Although both the fine-tuned ViTDet and the evaluation Faster-RCNN are trained on fastMRI+ annotations, this shared label source raises a proxy-validity concern, not a definitional reduction: no equation equates the optimized feature distance with the reported recall, and no parameter is fitted to the evaluation metric. The paper explicitly acknowledges the detector-proxy limitation in Section 4. Self-citations to the authors' prior work ([16], [17], [20]) are contextual or methodological and do not carry the central claim. Consequently, the paper's main results retain independent empirical content and no specific circular step can be exhibited.

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

The method depends on the standard MRI forward model, on the semantic meaningfulness of robust feature distances, on the reachability of pathologies within a small perturbation ball, and on the validity of detector-based clinical evaluation. The main free hyperparameters (r, Nrec, training budgets, proposal filters) are chosen without sensitivity analysis.

free parameters (4)

Perturbation radius r = 3
Hand-chosen to allow meaningful perturbations while keeping the number of new false positives low (Section 3.2); no sensitivity analysis is provided.
Number of SDR reconstructions Nrec = 3
Used in all experiments; no ablation of Nrec is reported.
Robust ViTDet adversarial training budget = r=10, 20,000 steps, batch size 16
Adopted from prior work [20] with the ℓ2 norm; its effect on downstream recall is not isolated.
Proposal filtering thresholds = keep 75% top RPN proposals; NMS IoU > 0.05; manual box jitter 75%-125% size, 25% position
These selection rules determine which regions are optimized; they are design choices without ablation.

assumptions (4)

standard math The MRI reconstruction forward model is y = A x with A = M F S, where S are coil sensitivity maps, F the Fourier transform, and M the undersampling mask.
Invoked in Eq. (1) and used to define data-consistent projections throughout Section 3.
domain assumption Euclidean distance in the feature space of an adversarially robust ViTDet is a meaningful semantic metric for pathology-relevant image differences.
The SDR objective in Eq. (3) assumes that maximizing this distance yields clinically meaningful diversity; the paper cites [6] but provides no direct validation for MRI pathology.
ad hoc to paper Alternative true pathologies are reachable by data-consistent perturbations within an ℓ2 ball of radius r=3 around the initial reconstruction.
The radius r=3 is chosen by hand; no evidence shows that missed pathologies lie in this ball or that the optimization can find them.
domain assumption A Faster-RCNN detector trained on fastMRI+ annotations is a valid proxy for human clinical diagnosis.
Section 4 replaces radiologist assessment with detector recall and mAP; the paper acknowledges this is a proxy.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mind the Detail: Uncovering Clinically Relevant Image Details in Accelerated MRI with Semantically Diverse Reconstructions." pith.science (2026). https://pith.science/paper/452MGXKL

@misc{pith2026250700670,
  author       = {Pith},
  title        = {Pith review of: Mind the Detail: Uncovering Clinically Relevant Image Details in Accelerated MRI with Semantically Diverse Reconstructions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/452MGXKL}},
  note         = {Machine review of arXiv:2507.00670}
}

read the original abstract

In recent years, accelerated MRI reconstruction based on deep learning has led to significant improvements in image quality with impressive results for high acceleration factors. However, from a clinical perspective image quality is only secondary; much more important is that all clinically relevant information is preserved in the reconstruction from heavily undersampled data. In this paper, we show that existing techniques, even when considering resampling for diffusion-based reconstruction, can fail to reconstruct small and rare pathologies, thus leading to potentially wrong diagnosis decisions (false negatives). To uncover the potentially missing clinical information we propose ``Semantically Diverse Reconstructions'' (\SDR), a method which, given an original reconstruction, generates novel reconstructions with enhanced semantic variability while all of them are fully consistent with the measured data. To evaluate \SDR automatically we train an object detector on the fastMRI+ dataset. We show that \SDR significantly reduces the chance of false-negative diagnoses (higher recall) and improves mean average precision compared to the original reconstructions. The code is available on https://github.com/NikolasMorshuis/SDR

Figures

Figures reproduced from arXiv: 2507.00670 by the authors.

**Figure 1.** Method overview. At high acceleration factors (12x shown here), naive MRI reconstruction suffers from artifacts (left). Recent reconstruction methods like DDIP [4] can recover image details and generate multiple reconstructions, but these often lack diversity, potentially leading to missed pathologies (top right). Our proposed SDR approach produces semantically diverse reconstructions, helping detect pathologies th… view at source ↗

**Figure 2.** Automatic Proposal Generation. Given an initial reconstruction, we filter the proposal boxes given by the Region Proposal Network (RPN) of the ViTDet to only keep the 75% of boxes with the highest score. We then apply Non-Maximum Suppression (NMS) on the remaining proposals. The remaining proposal boxes are then used for optimization with SDR-A. Init. Reconstruction Semantically Diverse Reconstructions All Detection… view at source ↗

**Figure 3.** Combining predictions of multiple reconstructions. We combine overlapping detection boxes and average their probability score across reconstructed images to compute the mAP of our detections. With Manual (SDR-M) Proposal Box Generation we address the scenario where a user suspects a pathology in a specific area and manually draws a bounding box. We simulate user annotations by perturbing the ground-truth boxes in … view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Evaluation of reconstruction methods. Both SDR-A and SDR-M outperform single and multiple reconstructions using the original reconstruction techniques (DDIP, CSGM, E2E-VN) in terms of recall (better detection of pathologies) while achieving similar or better mAP0.25. …

**Figure 5.** Figure 5: Qualitative Results The highlighted box in the ground truth shows a cartilage thickness loss (left) and a high grade ligament sprain (right). The initial reconstructions are often too smooth, preventing the detection of the pathologies. The semantic changes of SDR-M (l…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

PRIME-SVR: Physics-infoRmed Implicit Multi-Echo Slice-to-Volume Reconstruction for Fetal T2 mapping
physics.med-ph 2026-07 conditional novelty 6.0 of 10

A self-supervised, physics-regularized neural reconstruction produces high-resolution fetal brain T2 maps at 0.55 T and 1.5 T from multi-echo MRI, with reduced acquisition time.
CUTE-MRI: Conformalized Uncertainty-based framework for Time-adaptivE MRI
eess.IV 2025-08 reject novelty 6.0 of 10

An MRI acquisition framework that iteratively samples k-space and stops when a conformally calibrated uncertainty interval for a clinical metric meets a precision target.

Reference graph

Works this paper leans on

27 extracted references · 26 canonical work pages · cited by 2 Pith papers

[5]

In: ICLR (2024)

Cohen, N., Manor, H., Bahat, Y., Michaeli, T.: From posterior sampling to mean- ingful diversity in image restoration. In: ICLR (2024)

work page 2024
[1]

In: ECCV (2020)

Augustin, M., Meinke, A., Hein, M.: Adversarial robustness on in- and out- distribution improves explainability. In: ECCV (2020)

work page 2020
[2]

In: de Bruijne, M., Cattin, P.C., Cotin, S., Padoy, N., Speidel, S., Zheng, Y., Essert, C

Baumgartner, M., Jäger, P.F., Isensee, F., Maier-Hein, K.H.: nndetection: A self- configuring method for medical object detection. In: de Bruijne, M., Cattin, P.C., Cotin, S., Padoy, N., Speidel, S., Zheng, Y., Essert, C. (eds.) MICCAI 2021. pp. 530–539. Springer International Publishing, Cham (2021)

work page 2021
[3]

Medical Image Analysis p

Chung, H., Ye, J.C.: Score-based diffusion models for accelerated mri. Medical Image Analysis p. 102479 (2022)

work page 2022
[4]

In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G

Chung, H., Ye, J.C.: Deep diffusion image prior for efficient ood adaptation in 3d inverse problems. In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G. (eds.) ECCV 2024. pp. 432–455. Springer Nature Switzerland, Cham (2025)

work page 2025
[6]

In: IEEE SaTML (2025)

Croce, F., Schlarmann, C., Singh, N.D., Hein, M.: Adversarially robust clip models can induce better (robust) perceptual metrics. In: IEEE SaTML (2025)

work page 2025
[7]

In: NeurIPS Datasets and Benchmarks Track (2021)

Desai, A.D., Schmidt, A.M., Rubin, E.B., Sandino, C.M., Black, M.S., Mazzoli, V., Stevens, K.J., Boutin, R., Re, C., Gold, G.E., et al.: Skm-tea: A dataset for accelerated mri reconstruction with dense image labels for quantitative clinical evaluation. In: NeurIPS Datasets and Benchmarks Track (2021)

work page 2021
[8]

IJCV88, 303–338 (2010)

Everingham,M.,VanGool,L.,Williams,C.K.,Winn,J.,Zisserman,A.:Thepascal visual object classes (voc) challenge. IJCV88, 303–338 (2010)

work page 2010

Show all 27 references

[9]

In: WACV (2024)

Ganz, R., Elad, M.: Clipag: Towards generator-free text-to-image generation. In: WACV (2024)

2024
[10]

IEEE Access 11, 144349–144363 (2023)

Harman,F.,Selver,M.A.,Baris,M.M.,Canturk,A.,Oksuz,I.:Deeplearning-based meniscus tear detection from accelerated mri. IEEE Access 11, 144349–144363 (2023)

2023
[11]

NeurIPS (2021)

Jalal, A., Arvinte, M., Daras, G., Price, E., Dimakis, A.G., Tamir, J.I.: Robust compressed sensing mri with deep generative priors. NeurIPS (2021)

2021
[12]

Radiology: Artificial Intelligence 2(1), e190007 (2020), pMID: 32076662 10 Jan Nikolas Morshuis et al

Knoll, F., Zbontar, J., Sriram, A., Muckley, M.J., Bruno, M., Defazio, A., Par- ente, M., Geras, K.J., Katsnelson, J., Chandarana, H., Zhang, Z., Drozdzalv, M., Romero, A., Rabbat, M., Vincent, P., Pinkerton, J., Wang, D., Yakubova, N., Owens, E., Zitnick, C.L., Recht, M.P., S...

2020
[13]

Magnetic Resonance in Medicine92(1), 289– 302 (2024)

Küstner, T., Hammernik, K., Rueckert, D., Hepp, T., Gatidis, S.: Predictive un- certainty in deep learning–based mr image reconstruction using deep ensembles: Evaluation on the fastmri data set. Magnetic Resonance in Medicine92(1), 289– 302 (2024)

2024
[14]

In: ECCV

Li, Y., Mao, H., Girshick, R., He, K.: Exploring plain vision transformer backbones for object detection. In: ECCV. pp. 280–296. Springer (2022)

2022
[15]

In: CVPR (2015)

Mahendran, A., Vedaldi, A.: Understanding deep image representations by invert- ing them. In: CVPR (2015)

2015
[16]

In: Haq, N., Johnson, P., Maier, A., Qin, C., Würfl, T., Yoo, J

Morshuis, J.N., Gatidis, S., Hein, M., Baumgartner, C.F.: Adversarial robustness of mr image reconstruction under realistic perturbations. In: Haq, N., Johnson, P., Maier, A., Qin, C., Würfl, T., Yoo, J. (eds.) Machine Learning for Medical Image Reconstruction. pp. 24–33. Spri...

2022
[17]

In: Mukhopadhyay, A., Oksuz, I., Engelhardt, S., Mehrof, D., Yuan, Y

Morshuis,J.N.,Hein,M.,Baumgartner,C.F.:Segmentation-guidedmrireconstruc- tion for meaningfully diverse reconstructions. In: Mukhopadhyay, A., Oksuz, I., Engelhardt, S., Mehrof, D., Yuan, Y. (eds.) Deep Generative Models. pp. 180–190. Springer Nature Switzerland, Cham (2024)

2024
[18]

IEEE transactions on pattern analysis and machine intelligence 39(6), 1137–1149 (2016)

Ren, S., He, K., Girshick, R., Sun, J.: Faster r-cnn: Towards real-time object de- tection with region proposal networks. IEEE transactions on pattern analysis and machine intelligence 39(6), 1137–1149 (2016)

2016
[19]

In: NeurIPS (2019)

Santurkar, S., Ilyas, A., Tsipras, D., Engstrom, L., Tran, B., Madry, A.: Image synthesis with a single (robust) classifier. In: NeurIPS (2019)

2019
[20]

ICML (2024)

Schlarmann, C., Singh, N.D., Croce, F., Hein, M.: Robust clip: Unsupervised ad- versarial fine-tuning of vision embeddings for robust large vision-language models. ICML (2024)

2024
[21]

Radiology: Artificial Intelligence7(1), e240345 (2025)

Solomon, E., Johnson, P.M., Tan, Z., Tibrewala, R., Lui, Y.W., Knoll, F., Moy, L., Kim, S.G., Heacock, L.: fastmri breast: A publicly available radial k-space dataset of breast dynamic contrast-enhanced mri. Radiology: Artificial Intelligence7(1), e240345 (2025)

2025
[22]

In: Martel, A.L., Abolmaesumi, P., Stoyanov, D., Mateus, D., Zuluaga, M.A., Zhou, S.K., Racoceanu, D., Joskowicz, L

Sriram, A., Zbontar, J., Murrell, T., Defazio, A., Zitnick, C.L., Yakubova, N., Knoll, F., Johnson, P.: End-to-end variational networks for accelerated mri recon- struction. In: Martel, A.L., Abolmaesumi, P., Stoyanov, D., Mateus, D., Zuluaga, M.A., Zhou, S.K., Racoceanu, D., ...

2020
[23]

Scientific Data11(1), 404 (2024)

Tibrewala, R., Dutt, T., Tong, A., Ginocchio, L., Lattanzi, R., Keerthivasan, M.B., Baete, S.H., Chopra, S., Lui, Y.W., Sodickson, D.K., et al.: Fastmri prostate: A public, biparametric mri dataset to advance machine learning for prostate cancer imaging. Scientific Data11(1), ...

2024
[24]

In: ICLR (2019)

Tsipras, D., Santurkar, S., Engstrom, L., Turner, A., Madry, A.: Robustness may be at odds with accuracy. In: ICLR (2019)

2019
[25]

In: Wang, L., Dou, Q., Fletcher, P.T., Speidel, S., Li, S

Xie, Y., Li, Q.: Measurement-conditioned denoising diffusion probabilistic model for under-sampled medical image reconstruction. In: Wang, L., Dou, Q., Fletcher, P.T., Speidel, S., Li, S. (eds.) MICCAI 2022. pp. 655–664. Springer Nature Switzer- land, Cham (2022)

2022
[26]

Yen, C.Y., Singhal, R., Sharma, U., Ranganath, R., Chopra, S., Pinto, L.: Adaptive samplingofk-spaceinmagneticresonanceforrapidpathologyprediction.In:ICML (2024)

2024
[27]

Scientific Data 9(1), 152 (2022)

Zhao, R., Yaman, B., Zhang, Y., Stewart, R., Dixon, A., Knoll, F., Huang, Z., Lui, Y.W., Hansen, M.S., Lungren, M.P.: fastmri+, clinical pathology annotations for knee and brain fully sampled magnetic resonance imaging data. Scientific Data 9(1), 152 (2022)

2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [5]

In: ICLR (2024)

Cohen, N., Manor, H., Bahat, Y., Michaeli, T.: From posterior sampling to mean- ingful diversity in image restoration. In: ICLR (2024)

work page 2024

[2] [1]

In: ECCV (2020)

Augustin, M., Meinke, A., Hein, M.: Adversarial robustness on in- and out- distribution improves explainability. In: ECCV (2020)

work page 2020

[3] [2]

In: de Bruijne, M., Cattin, P.C., Cotin, S., Padoy, N., Speidel, S., Zheng, Y., Essert, C

Baumgartner, M., Jäger, P.F., Isensee, F., Maier-Hein, K.H.: nndetection: A self- configuring method for medical object detection. In: de Bruijne, M., Cattin, P.C., Cotin, S., Padoy, N., Speidel, S., Zheng, Y., Essert, C. (eds.) MICCAI 2021. pp. 530–539. Springer International Publishing, Cham (2021)

work page 2021

[4] [3]

Medical Image Analysis p

Chung, H., Ye, J.C.: Score-based diffusion models for accelerated mri. Medical Image Analysis p. 102479 (2022)

work page 2022

[5] [4]

In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G

Chung, H., Ye, J.C.: Deep diffusion image prior for efficient ood adaptation in 3d inverse problems. In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G. (eds.) ECCV 2024. pp. 432–455. Springer Nature Switzerland, Cham (2025)

work page 2025

[6] [6]

In: IEEE SaTML (2025)

Croce, F., Schlarmann, C., Singh, N.D., Hein, M.: Adversarially robust clip models can induce better (robust) perceptual metrics. In: IEEE SaTML (2025)

work page 2025

[7] [7]

In: NeurIPS Datasets and Benchmarks Track (2021)

Desai, A.D., Schmidt, A.M., Rubin, E.B., Sandino, C.M., Black, M.S., Mazzoli, V., Stevens, K.J., Boutin, R., Re, C., Gold, G.E., et al.: Skm-tea: A dataset for accelerated mri reconstruction with dense image labels for quantitative clinical evaluation. In: NeurIPS Datasets and Benchmarks Track (2021)

work page 2021

[8] [8]

IJCV88, 303–338 (2010)

Everingham,M.,VanGool,L.,Williams,C.K.,Winn,J.,Zisserman,A.:Thepascal visual object classes (voc) challenge. IJCV88, 303–338 (2010)

work page 2010