List of Papers Browse by Subject Areas Author List
Abstract
Multi-modal Magnetic Resonance Imaging (MRI) translation leverages information from source MRI sequences to generate target modalities, enabling comprehensive diagnosis while overcoming the limitations of acquiring all sequences. While existing deep-learning-based multi-modal MRI translation methods have shown promising potential, they still face two key challenges: 1) lack of reliable uncertainty quantification for synthesized images, and 2) limited robustness when deployed across different medical centers. To address these challenges, we propose a novel framework that reformulates multi-modal MRI translation as a multi-modal evidential regression problem with distribution calibration. Our approach incorporates two key components: 1) an evidential regression module that estimates uncertainties from different source modalities and an explicit distribution mixture strategy for transparent multi-modal fusion, and 2) a distribution calibration mechanism that adapts to source-target mapping shifts to ensure consistent performance across different medical centers. Extensive experiments on three datasets from the BraTS2023 challenge demonstrate that our framework achieves superior performance and robustness across domains.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/0738_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{LiuJiy_Multimodal_MICCAI2025,
author = { Liu, Jiyao and Gao, Shangqi and Li, Yuxin and Liu, Lihao and Gao, Xin and Xing, Zhaohu and Ning, Junzhi and Su, Yanzhou and Zhang, Xiao-Yong and He, Junjun and Xu, Ningsheng and Zhuang, Xiahai},
title = { { Multi-modal MRI Translation via Evidential Regression and Distribution Calibration } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15967},
month = {September},
page = {363 -- 373}
}
Reviews
Review #1
- Please describe the contribution of the paper
Uncertainty Quantification: The proposed framework provides a novel means of estimating uncertainty in multi-modal MRI synthesis through the Mixture of Normal-Inverse Gamma (MoNIG) strategy and evidential regression. This allows medically relevant confidence measures for synthesized images.
Domain Adaptation: Introducing distribution calibration mechanisms ensures robust performance across different medical scanners and institutions.
Comprehensive Validation: Experiments on datasets of BraTS2023, such as BraSyn, BraTS-Africa, and BraTS-PED, demonstrate its ability to yield superior results, even in cross-center scenarios.
Innovative Probabilistic Approach: Reframes MRI translation into a probabilistic framework, leveraging statistical and deep learning methods for clinically interpretable outcomes.
- 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.
Critical Problem Solving: Effectively addresses two key issues facing MRI translation—uncertainty quantification and domain robustness.
Practical Relevance: Directly supports clinical scenarios by improving the utility and reliability of synthesized images.
Robust Experimentation: Demonstrates superior quantitative performance on multiple datasets while validating adaptability in diverse environments.
Innovative Methodology: Combines advanced statistical techniques with deep learning for robust image fusion in multiple domains.
- 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.
Computational Complexity: The complexity of model training and inference could hinder its deployment in real-time clinical environments.
Limited Experimental Context: More extensive datasets and diverse clinical cases are needed to generalize the framework’s robustness.
Unaddressed Scalability: No explicit strategies for reducing the need for annotated data or simplifying the model complexity are outlined, posing challenges for broad application.
- 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 provide sufficient information for 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
The reviewer is curious about a larger-scale mult-center dataset to validate the proposed method. Frankly speaking, generating one MRI modality from another modality is challenging and sometimes the generated scan is not reliable for clinical purpose (such as for tumor diagnosis).
- 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?
The organization and writing of this submission as well as the interesting idea in the manuscript for MRI translation are the major factors.
- 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 paper makes two main contributions. First it reinterprets MRI translation as an evidential regression problem, meaning the network predicts a probability distribution for each voxel. This approach supplies explicit uncertainty estimates, helping clinicians judge whether a synthesized image region is trustworthy. Secondly, it introduces a calibration step that adjusts the learned distribution to new clinical environments with minimal extra data. This mechanism helps the model remain robust under varied scanning protocols and patient populations.
- 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 authors recast multi-modal MRI translation as an evidential regression task via a NIG model, rather than a simple L1/L2 regression. By doing so, they explicitly learn predictive distributions and thereby quantify uncertainty at each voxel. This is interesting because most medical image synthesis approaches produce only a point estimate without transparent reliability measures.
- They fuse multiple source modalities through a MoNIG rule. This explicit fusion strategy incorporates both epistemic and aleatoric uncertainties. It is important clinically, because it highlights where a synthesized image can be trusted, a capability not typically offered by standard GAN- or diffusion-based approaches.
- The paper introduces a quantile-regression-based calibration step. This is unusual because most image-translation research does not deal with calibration or distribution shifts at deployment. By adapting to new clinical centers with minimal data, the approach addresses a common but underexplored barrier for real-world multi-modal MRI synthesis.
- 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 core idea of employing Normal–Inverse Gamma for evidential regression closely follows Amini et al. [1] and Sensoy et al. [22]. Likewise, the MoNIG strategy for multi-view fusion was proposed in Ma et al. [18]. While the authors apply these methods to multi-modal MRI translation (which is helpful), the fundamental probabilistic modeling and summation rules are not newly invented here.
- Although the paper tests across BraSyn, BraTS-Africa, and BraTS-PED, these remain relatively homogeneous datasets—mostly variations of glioma MRI. Broader external validation is not presented.
- 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?
The paper brings together an evidential regression approach, a clear multi-modal fusion mechanism (MoNIG), and a simple calibration procedure to tackle domain shifts. Although some evidential aspects and the MoNIG approach draw from established methodologies, this is still a well-crafted application for MRI synthesis, with robust experiments on multiple relevant datasets.
- 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 authors introduce a regression-based MRI translation method that incorporates uncertainty quantification. Their approach integrates a distribution calibration strategy using quantile regression to improve generalization across datasets with minimal fine-tuning. The method is evaluated on the BraSyn, BraTS-Africa, and BraTS-PED datasets, which serve as cross-center and cross-population test sets. The proposed method outperforms baseline approaches on nearly all evaluated metrics.
- 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.
This paper addresses two key challenges in MRI image translation and medical imaging more broadly: uncertainty quantification and generalizability across data from different medical centres. The work is well-motivated and clearly written, with a strong methodological foundation. The proposed approach appears technically sound, and the experimental results are compelling, demonstrating consistent improvements over baselines across multiple datasets.
- 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.
While the proposed method is promising, the exploration of the uncertainty modelling component is somewhat limited. It would be helpful if the authors clarified what Figure 4 is showing, either in the main text or the figure caption. A quantitative evaluation of the uncertainty modelling would have been nice to see, although I appreciate that space is limited. It would also have been nice to see the qualitative results in Fig 3 in a different plane as well as axial. Some aspects of the methods and results sections could also be clearer. For instance, a brief sentence in Section 2.3 linking the parameter maps to the distributions introduced in Section 2.1 would help readers better understand how the translated images are generated. Additionally, the sentence “We conducted experiments using T1n and T2w as input to synthesize T2f and T1c modalities. The results show the average performance across T2f and T1c synthesis,” which currently appears only in the table title, should be moved into the main text for clarity.
- 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
Could the authors please clarify what is the purpose of the encoder? Why are the parameter maps not learnt from the generator directly?
- 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?
This paper is well-motivated and presents an interesting approach to image translation while modelling uncertainty, and effectively accounts for site differences. The proposed method demonstrates strong performance outperforming baselines in nearly all metrics. Some clarifications needed but overall the paper is solid.
- 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
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- 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