List of Papers Browse by Subject Areas Author List
Abstract
Synthesizing high-fidelity contrast-enhanced MRI from non-contrast sequences is clinically valuable for safer and more efficient breast cancer screening, yet remains challenging due to complex lesion textures and heterogeneous enhancement patterns. In this work, we propose SAFE-Diff, a multi-scale attention–enhanced diffusion network for contrast-agent-free breast MRI synthesis. The framework combines convolutional feature extraction with complementary attention mechanisms across spatial scales, while a window-based self-attention module inspired by Swin Transformers is employed at the bottleneck to capture long-range anatomical context via shifted-window interactions. To improve robustness in heterogeneous tissues, we introduce an uncertainty-aware learning formulation that adaptively reweights the reconstruction objective based on spatially varying prediction confidence. In addition, a feature-dispersive learning strategy is applied to intermediate latent representations to mitigate representation redundancy and preserve fine-grained diagnostic details. Extensive experiments demonstrate that SAFE-Diff accurately reconstructs contrast-enhancement characteristics and outperforms state-of-the-art approaches. Code is available at https://github.com/Netherlands-Cancer-Institute/SAFE-Diff.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/3480_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/Netherlands-Cancer-Institute/SAFE-Diff
Link to the Dataset(s)
N/A
BibTex
@InProceedings{ZhaTia_SAFEDiff_MICCAI2026,
author = { Zhang, Tianyu AND Liang, Xinglong AND van Dijk, Jarek AND Han, Luyi AND Lu, Chunyao AND Portaluri, Antonio AND Xie, Xinghe AND Duan, Yaofei AND Rasoolzadeh, Nika AND Wang, Xin AND Gao, Yuan AND He, Muzhen AND Sun, Yue AND Teuwen, Jonas AND Tan, Tao AND Mann, Ritse},
title = { { SAFE-Diff: Scale-Aware Attention and Feature-Dispersive Diffusion with Uncertainty Estimation for Contrast-Enhanced Breast MRI Synthesis } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16890},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper proposes a SAFE-Diff framework for the contrast-enhanced breast MRI synthesis task. It proposes the window-based self-attention module inspired by Swin Transformers to capture long-range anatomical context, and uncertainty-aware learning to improve robustness, and feature-dispersive learning to mitigate redundancy.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The virtual enhancement from non-contrast MRI and DWI is meaningful. Using an external dataset for generalization evaluation. The details of the experiments are clear.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.The proposed method appears to be a mere combination of multiple existing innovative modules. The authors do not provide motivation for employing these modules, and the differences and novelties between the proposed modules and existing ones are not clearly demonstrated.
2.The use of DWI, whether it is adopted and how it is integrated, is unclear. The authors describe the role of DWI in the third paragraph, yet the specific way DWI is “combined with the stable anatomical context of pre-contrast T1” is not explained in the method. Is it directly concatenated as the conditional input c?
3.Both the caption and the design of Figure 1 are highly unclear. No detailed interpretation is given for each subfigure. What does “image manifold” stand for here, and why is it introduced?
4.Section 2 is poorly organized. The description of data and the presentation of the method should be separated, and the method section lacks clear focus.
5.Visualization results and analyses for the comparative methods are missing.
6.Why is x₀-prediction adopted instead of the commonly used noise-prediction or v-prediction?
- Please rate the clarity and organization of this paper
Poor
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper is somewhat lacking in methodological design, overall structure, and experimental analysis, resulting in low completeness of the presented work.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
This paper presents SAFE-Diff for Contrast-Enhanced Breast MRI Synthesis. Compared to a standard canonical diffusion framework, this paper introduced: 1.heteroscedastic uncertainty-aware formulation; 2.feature-dispersive regularization; 3.direct x0 estimation with multi-scale attention; 4.perceptual consistency loss.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The use of those components are relatively novel for breast contrast generation application. The visualization of uncertainty maps shows that it is actually doing what it is supposed to do. The author conducted sufficient experiments and ablations, and the experiments showed sufficient improvements in different metrics.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
There’s no explicit reasoning for why the authors choose to include feature dispersive learning strategy at the bottleneck and the final decoder stage. Lack of downstream evaluations. I just want to say that metrics such as SSIM & PSNR & NMSE and nHFEN are not the best quantifier for potentially how useful the generated images are diagnosis.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This paper addressed a clinically important problem and also showed some satisifying results in metrics. The experiments are on in house and external dataset. The framework is fair and the ablations supports the idea.
- Reviewer confidence
Somewhat confident (2)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
The authors propose a swin transformer based diffusion network for synthesis of DCE breast MRI images from non-contrast images. The authors also propose an uncertainty aware learning method where the objective function is dynamically reweighted based on the spatially varying model prediction confidence across different.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.Problem formulation and the choice of different modules are explained well 2.Feature dispersive learning strategy and mask aware perceptual loss are novel contributions of this work 3.Ablation study justifies the need for the different components of the proposed solution
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.How were the anatomy masks for the mask-aware perceptual loss derived? Expert drawn or automatically segmented using an algorithm? This information is missing in the description 2.Pixel-based quantitative metrics (PSNR, SSIM, NMSE) do not give a full picture of the model performance. Even though the authors have reported these metrics inside the tumor region, the values are significantly lower when compared to the global metrics. Given that they have localized masks of tumor, breast and background regions for each case, the authors should report lesion based metrics like Contrast-to-noise Ratio (CNR), Contrast Enhancement Percentage (CEP) and Lesion-to-background Ratio (LBR). 3.For each image in the qualitative figure, the authors should display the enhancement kinetic curve (washout pattern) and a dot plot overlaid with the average signal intensities of the tumor region of the real-Gad and synth-Gad images. 4.The authors should try providing at least one example for each different washout pattern thus highlighting the model performance on different types of tumors.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
1.Does each case have only one post-contrast T1w sequence, or multiple? This needs to be clarified. If the latter is true, which sequence is chosen as the post-contrast sequence? 2.Are all images in the dataset perfectly aligned? The authors should add discussion points on how the model tackles difference in spatial alignment. 3.What are the failure modes? The authors should show a few examples of where the model fails and discuss the reasons.
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper has some novel methodological contributions and addresses an important problem in breast MRI. The paper should be accepted if the authors satisfactorily address the changes/comments that are mentioned above - especially regarding the lesion specific quantitative metrics.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
- Lack of downstream task evaluations.
- Clarify a few implementation details.
- Improve paper organization to improve readability.