List of Papers Browse by Subject Areas Author List
Abstract
Reducing MRI scan times can improve patient care and lower healthcare costs. Many acceleration methods are designed to reconstruct diagnostic-quality images from sparse k-space data, via an ill-posed or ill-conditioned linear inverse problem (LIP). To address the resulting ambiguities, it is crucial to incorporate prior knowledge into the optimization problem, e.g., in the form of regularization. Another form of prior knowledge less commonly used in medical imaging is the side information obtained from sources other than the current acquisition. In this paper, we present the Trust-Guided Variational Network (TGVN), an end-to-end deep learning framework that effectively and reliably integrates side information into LIPs. We demonstrate its effectiveness in multi-coil, multi-contrast MRI reconstruction, where incomplete or low-SNR measurements from one contrast are used as side information to reconstruct high-quality images of another contrast from heavily undersampled data. TGVN is robust across different contrasts, anatomies, and field strengths. Compared to baselines utilizing side information, TGVN achieves superior image quality while preserving subtle pathological features even at challenging acceleration levels, drastically speeding up acquisition while minimizing hallucinations. Source code and dataset splits are available on github.com/sodicksonlab/TGVN.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/4945_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/sodicksonlab/TGVN
Link to the Dataset(s)
https://fastmri.med.nyu.edu
https://zenodo.org/records/8056074
BibTex
@InProceedings{AtaArd_Harnessing_MICCAI2025,
author = { Atalık, Arda and Chopra, Sumit and Sodickson, Daniel K.},
title = { { Harnessing Side Information for Highly Accelerated MRI } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15972},
month = {September},
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper proposed a new method to leverage non-image (and also image-based) information for accelerated MRI reconstruction.
- 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 method is theoretically well-grounded and effectively incorporates side information via a learnable module. 2.The manuscript is easy to follow, with a clear logical flow. 3.The experiments were conducted on multiple public datasets. TGVN includes ablation studies to evaluate the effectiveness of extracting useful side information from under-sampled k-space in multi-contrast settings, comparing performance with and without side information.
- 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.
-
This paper lacks clarity on the domain of sharing and overlooks comparisons with diffusion-based methods [1][2].
-
TGVN should be compared with [2], as both studies use VarNet as the main framework.
[1] Li, Guangyuan, et al. “Rethinking diffusion model for multi-contrast mri super-resolution.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024. [2] Xuan, Kai, et al. “Multimodal MRI reconstruction assisted with spatial alignment network.” IEEE Transactions on Medical Imaging 41.9 (2022): 2499-2509.
-
- 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.
- 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?
lacking novelty and strong comparative experiments.
- 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
This paper proposes a new method for performing MRI reconstruction with side information.
- 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 results are promising, and the methods are based on good mathematical insights into the structure of the inverse problem. Unlike many submissions to MICCAI this year, this paper appears to use realistic simulations of accelerated MRI 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.
The paper claims that using side information for medical image reconstruction is “less common”. The literature review cites References [2,9,5] as “earlier attempts that utilized handcrafted priors”. These descriptions are misleading, because there is actually a substantial amount of work on this topic, including work that is substantially earlier than [2,9,5]. Take a look at Chapter 5.7 (“Reconstruction using side information”) of the book DOI:10.1016/B978-0-12-822726-8.00014-2 There are many earlier papers on this topic from the 1980s, 1990s, and 2000s, as well as many recent papers on this subject.
- 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?
I’d like the paper to be revised to avoid misleading statements that hide the substantial amount of previous work that has been done on this topic. The methods and results themselves are interesting.
- 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 #3
- Please describe the contribution of the paper
The paper proposes to modify iterative MRI reconstruction such that it incorporates side information from other contrasts using an approximate orthogonal operator and a learned model. This method combines with “traditional” iterative reconstruction using deep learning to improve reconstruction quality.
- 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 paper is well-written, without excessive equations, and the idea is presented clearly.
- 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 experimental investigation is incomplete - it is not clear how well would the method performed compared to a single monolith neural network that takes as inputs multiple contrasts (measurements from side channel; and current recon from target contrasts) and outputs just the target one. Clarifying this via a direct experiment would greatly improve confidence in the method and its novelty.
- 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.
- 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
- While the idea of including “other” types of side information such as “Electronic Health Records” in the bottom of Figure 1 is appealing, the method presented in the paper is applicable only to other contrast side information, and it would be better if the figure was revised to remain factual to the contents of the paper. It is not clear how text or other features would be “projected onto the ambiguous space of 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?
Clear methodology and commitment to reproducibility (promised source code release).
- 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
We thank Reviewer #1 for acknowledging the theoretical grounding of our work, the clarity of our paper, and our use of informative ablation studies (6.1, 6.2, 6.3). On comparisons to [1, 2] (7.1, 7.2): [1] addresses a different problem (super-resolution), and direct comparisons would be difficult to quantify. The novelty in [2] lies in handling spatial alignment via another module, not in terms of using side information. Instead, we included comparisons with three well-established baselines that are designed for reconstruction of undersampled data and that directly address MRI reconstruction with side information. These baselines are also more recent than [2] and were published either in the same venue as [2] or in venues of comparable quality. Also, we do not overlook comparisons with diffusion-based methods, as one of our three baselines, DMSI, is in fact a diffusion-based method which was designed to incorporate side information into MRI reconstruction. On novelty and strong comparative experiments (12): Our contribution lies not only in leveraging side information, but in how we enforce it using the physics of the forward operator to ensure that the guidance of side information stays data consistent. This differs from existing architectures in both formulation and motivation, introducing a novel, physically grounded mechanism for integrating side information in a data-consistent manner. We compared our method against strong, recently published baselines, each designed to leverage side information in MRI reconstruction.
We thank Reviewer #2 for their feedback. We are especially grateful for their recognition of the mathematical insights underlying our method and our use of realistic simulations for accelerated MRI. On the novelty of using side information (7, 12): We agree that the topic of image reconstruction using side information has a long and rich history, with both early foundational work and more recent developments. Our intention was not to understate prior contributions, but rather to emphasize that the use of population-based priors has been far more commonly explored than the use of patient-specific priors. Due to space constraints, we cited only a few representative works. Based on your feedback, we will gladly revise and expand the discussion to better reflect prior works on image reconstruction with side information, e.g., 10.1109/23.173225, 10.1109/42.251117, 10.1109/42.538945, 10.1002/mrm.21536, and 10.1109/TMI.2010.2046673. On reproducibility (9): We would like to clarify that we have committed to releasing the source code upon acceptance, as stated in the final sentence of the abstract.
We thank Reviewer #3 for the constructive feedback and comments on the clarity of our presentation. On comparison with a monolithic neural network baseline (7): We agree that a unified network using both main and side information as joint inputs is a natural baseline. We previously implemented a U-Net that takes multi-coil, complex-valued images from both sources and outputs the RSS target. However, the results were not competitive with the physics-based baselines presented in the manuscript, which better preserve data consistency. Though we already have the U-Net baseline results, we want to be careful not to violate MICCAI’s rule against including new experimental results unless explicitly permitted. In any case, we can confirm that we have in fact considered your suggested comparison. On alternative types of side information (10): We included a brief discussion in the conclusion on how non-image features could be incorporated into TGVN via straightforward modifications to the learnable H block. In brief, one can replace the image encoder in the H block with another encoder tailored to the side information in question, or else one can use representations from pre-trained encoders as input into the H block decoder. We will revise the presentation to clarify this point both in text and the accompanying figure.
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’
N/A