List of Papers Browse by Subject Areas Author List
Abstract
Preclinical imaging studies are vital to the research, development, and evaluation of new medical therapies. Images acquired during these studies often have high in-plane resolution but low through-plane resolution, resulting in highly anisotropic volumes that hamper downstream volumetric analysis. Additionally, since there are no image acquisition standards across studies, limited general training data restricts the ability of conventional supervised learning approaches that aim to improve resolution. In this work, we compare two super-resolution (SR) methods that do not require additional training data. The first is a self-SR approach that learns from in-plane patches without relying on external data. The second is a denoising diffusion probabilistic model trained solely to represent high-resolution rat magnetic resonance images, and SR is based on the inference step proposed by denoising diffusion null-space models (DDNM). We name this second approach Biplanar DDNM Averaging (BiDA). We evaluate both methods at two scale factors (2.5x and 5x) in withheld data with regard to signal recovery and downstream task performance. We further evaluate these methods when applied to mice instead of rats. Both methods experimentally resulted in good signal recovery performance, but only the images super-resolved by BiDA were accurately skullstripped downstream. While performance was good in rats, BiDA did not generalize satisfactorily to mice, which have smaller heads.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2076_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{GhaOma_Exploring_MICCAI2025,
author = { Gharib, Omar A. M. and Remedios, Samuel W. and Dewey, Blake E. and Prince, Jerry L. and Carass, Aaron},
title = { { Exploring the feasibility of zero-shot super-resolution in preclinical imaging } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15961},
month = {September},
page = {185 -- 194}
}
Reviews
Review #1
- Please describe the contribution of the paper
This manuscript presents four key contributions: 1) curating and preprocessing high-resolution 2D rat imaging data to train a denoising diffusion probabilistic model (DDPM) that approximates this data distribution; 2) proposing BiDA, a novel method leveraging denoising diffusion null-space models (DDNMs) for volumetric super-resolution (SR); 3) evaluating BiDA against other state-of-the-art SR methods on withheld rat data, including assessments of signal recovery and performance on downstream tasks; and 4) testing BiDA’s generalization to mouse imaging data, highlighting potential limitations in cross-species applicability.
- 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.
- Theoretical soundness: The method is theoretically well-grounded, leveraging the strong generalization capability of DDNM in a zero-shot setting to perform super-resolution on two planes, followed by fusion of the results.
- Comprehensive experiments: The experimental validation is thorough, including tests on out-of-domain datasets, which is a notable strength. In addition, the method’s effectiveness is evaluated through downstream segmentation tasks, where the improved segmentation performance demonstrates its practical utility and potential to benefit segmentation models.
- Clear writing and logical structure: The manuscript is well-organized and easy to follow, with a clear logical flow.
- Clinical relevance: The proposed method has potential clinical value, as it enhances volumetric resolution and can support downstream analyses such as segmentation.
- 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.
-
The paper does not provide an analysis of using only axial or sagittal DDNM individually for slice-wise reconstruction. Without such comparisons, it is unclear whether the proposed dual-branch (biplanar) approach is truly effective or necessary.
-
The novelty is not very strong. For example, DiffusionMBIR can achieve 3D reconstruction by leveraging 2D priors, simply by modifying the estimation of operator A with a 3D kernel. It remains unclear what advantages the proposed approach—reconstructing from two orthogonal planes—offers over such existing methods.
-
- Please rate the clarity and organization of this paper
Satisfactory
- 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 submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure reproducibility.
- 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/2025/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 — could be rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
based on novelty and experimental results.
- Reviewer confidence
Confident but not absolutely certain (3)
- [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
The BiDA method for supersolution of preclinical data that also works for out-of-domain data.
- 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) The BiDA method combines axial and sagittal DDPM methods to provide final hgih resolution methods. 2) The evaluation and comparison of the methods with in-domain and out-of-domain data.
- 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) Lack of details for the Self-SR method. 2) The defintion of the operator A in Eq. (1) is not introduced. 3) The Self-SR method has better PSNR than the proposed BiDA method.
- 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 submission has provided an anonymized link to the source code, dataset, or any other dependencies.
- 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/2025/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.
(4) Weak Accept — could be accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Some key information of the method is missing. But the method and results are useful to super resolution in preclinical imaging.
- Reviewer confidence
Confident but not absolutely certain (3)
- [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 paper introduces Biplanar DDNM Averaging (BiDA), a zero-shot super-resolution (SR) method for preclinical imaging, specifically addressing the challenge of anisotropic MRI volumes in rodent studies. BiDA leverages two orthogonal 2D denoising diffusion probabilistic models (DDPMs) trained on high-resolution rat MRI data and applies denoising diffusion null-space model (DDNM) inference to perform volumetric SR. The method is evaluated on both in-domain (rat) and out-of-domain (mouse) data, demonstrating strong signal recovery and improved downstream task performance (e.g., skull stripping) compared to traditional interpolation and self-SR methods. The work highlights the potential of diffusion-based models for zero-shot SR in biomedical imaging while also identifying limitations in cross-species generalization.
- 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) The introduction of new architecture, which combines two orthogonal 2D DDPMs with DDNM inference for volumetric SR, is a significant contribution. The biplanar averaging approach is innovative and addresses the challenge of anisotropic resolution in preclinical MRI. 2) The paper demonstrates the feasibility of zero-shot SR, which is particularly valuable in preclinical imaging where training data is often limited or domain-specific. 3) The study includes both quantitative (PSNR, cDSC) and qualitative evaluations across multiple datasets, including in-domain and out-of-domain scenarios
- 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 method performs well on rat data but shows limitations when applied to mice, which have smaller heads. This suggests a need for species-specific adaptations or further model tuning. 2) While the paper compares BiDA to cubic interpolation and self-SR, it could benefit from a more extensive comparison to other recent zero-shot or diffusion-based SR methods to better contextualize its contributions. 3) The paper does not discuss the computational cost or runtime of BiDA, which could be a practical concern for large-scale preclinical studies. 4) The reliance on 2D slice-wise processing may introduce artifacts or inconsistencies in the volumetric output. A fully 3D diffusion model could potentially improve spatial coherence.
- 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 submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure reproducibility.
- 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/2025/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.
(4) Weak Accept — could be accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
While this work is not without problems and there is room for improvement, these could all realistically be fixed.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
The authors addressed all concerns regarding their work.
Author Feedback
We thank the reviewers for their helpful comments and insights and address them below. R1C1: Separate axial or sagittal DDNM analysis to clarify the benefits of the proposed dual-branch (biplanar) approach is necessary. A: Slice-to-slice inconsistencies stemming from 2D processing of volumes is well-known in the community (for instance, see R3C4). As such, and because of the space constraints, we did not include this result in our experiments. However, we will include relevant citations in the introduction to clarify this point.
R1C2: Novelty and comparison to DiffusionMBIR. A: We thank the reviewer for bringing DiffusionMBIR to our attention. While it would serve as an excellent comparison in a follow-up journal article, the core novelty of our submission lies in demonstrating, for the first time, the efficacy of zero-shot super-resolution via DDNM with 2D priors in preclinical imaging. To our knowledge, no prior work has evaluated such an approach in this domain. We will emphasize this point by underlining these contributions in the introduction.
R2C1: Lack of details for the Self-SR method. A: We thank the reviewer for their observation. While we did cite the Self-SR method in Sec. 4, we agree that it should be introduced earlier for a clearer exposition. Towards this, we will introduce and reference Self-SR earlier, in Sec. 3, so it is clear to the reader what methods are involved in this study.
R2C2: The operator A in Eq. (1) is not introduced. A: We defined A in the first paragraph of Sec. 2: Methods.
R2C3: Self-SR has better PSNR than the proposed BiDA method. A: This is expected according to the perception-distortion trade-off, which we addressed in the 3rd paragraph of Sec. 3: Experiments and Results.
R3C1: BiDA performs well on rat data but shows limitations when applied to mice, which have smaller heads. This suggests a need for species-specific adaptations or further model tuning. A: We agree, and commented on this in paragraph 2 in Sec. 4: Conclusion.
R3C2: While the paper compares BiDA to cubic interpolation and self-SR, it could benefit from a more extensive comparison to other recent zero-shot or diffusion-based SR methods to better contextualize its contributions. A: We agree that extensions to this work should incorporate further comparisons, such as DiffusionMBIR and CSGM. Such comparisons will be included in follow-up work.
R3C3: Computational cost or runtime of BiDA A: We thank the reviewer for noticing our oversight. We will include a comment on the computational cost in paragraph 7 of Sec. 3: Experiments and Results at the bottom of pg. 7 (which we have checked will fit). Specifically, we will add: On an NVIDIA A16, BiDA processes 144×144×64 in ~10s (<10 GB vRAM); Self-SR takes ~5 min at similar memory.
R3C4: The reliance on 2D slice-wise processing may introduce artifacts or inconsistencies in the volumetric output. A fully 3D diffusion model could potentially improve spatial coherence. A: We agree. This is why we proposed BiDA, which alongside DDNM yields volumes that are consistent with the LR volume while also not exhibiting slice-wise artifacts. Furthermore, a full 3D model is not feasible for preclinical data for which 3D images are rarely acquired. This comment is addressed in the first paragraph of Sec. 1: Introduction, in the original manuscript.
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.
N/A
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’
N/A
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’
N/A
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’
The rebuttal have covered the key issues raised by all the reviewers.