List of Papers Browse by Subject Areas Author List
Abstract
Fetal cortical surface reconstruction is crucial for quantitative analysis of normal and abnormal prenatal brain development. While there are many cortical surface reconstruction methods available for adults and infants, there remains a notable scarcity of dedicated techniques for fetal cortical surface reconstruction. Of note, fetal brain MR images present unique challenges, characterized by nonuniform low tissue contrast associated with extremely rapid brain development and folding during the prenatal stages and low imaging resolution, as well as susceptibility to severe motion artifacts. Moreover, the smaller size of fetal brains results in much narrower cortical ribbons and sulci. Consequently, the fetal cortical surfaces are more prone to be influenced by partial volume effects and tissue boundary ambiguities. In this work, we develop a multi-task, priori-knowledge supervised fetal cortical surface reconstruction method based on deep learning. Our method incorporates a cycle-consistent strategy, utilizing prior knowledge and multiple stationary velocity fields to enhance its representation capabilities, enabling effective learning of diffeomorphic deformations from the template surface mesh to the inner and outer surfaces. Specifically, our framework involves iteratively refining both inner and outer surfaces in a cyclical manner by mutually guiding each other, thus improving accuracy especially for ambiguous and challenging cortical regions. Evaluation on a fetal MRI dataset with 83 subjects shows the superiority of our method with a geometric error of 0.229 ± 0.047 mm and 0.023 ± 0.058% self-intersecting faces, indicating promising surface geometric and topological accuracy. These results demonstrate a great advancement over state-of-the-art deep learning methods, while maintaining high computational efficiency.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2024/paper/3476_paper.pdf
SharedIt Link: pending
SpringerLink (DOI): pending
Supplementary Material: N/A
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{Don_Cycleconsistent_MICCAI2024,
author = { Dong, Xiuyu and Wu, Zhengwang and Ma, Laifa and Wang, Ya and Tang, Kaibo and Zhang, He and Lin, Weili and Li, Gang},
title = { { Cycle-consistent Learning for Fetal Cortical Surface Reconstruction } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
year = {2024},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15007},
month = {October},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper presents a multi-task, priori-knowledge supervised method for fetal cortical surface reconstruction. Experiments on a fetuses dataset (21-35 weeks) shows the superiority of the proposed method.
- Please list the main strengths of the paper; you should write about 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.
An important problem was being identified and addressed and the introduction offers a comprehensive overview and explanation of the paper’s contributions.
- Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
-
The number of iterations in this work is 2, and it raises concerns about the convergence since 2 is a very small number. It would be better if the authors could conduct an ablation study on the iteration numbers.
-
For the visual results shown in Fig.3, the reconstruction results from the proposed method do not show significant improvement compared to the results from CoTAN. It would be better if the authors could highlight the differences in the figure.
-
How does the pretraining strategy affect the performance of the proposed method? It would be better if the authors could conduct an ablation study on this.
-
- Please rate the clarity and organization of this paper
Excellent
- 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.
- Do you have any additional comments regarding the paper’s reproducibility?
The pre-training dataset is public dataset from dHCP, and the author did not mention whether the code and the train fetal data will be released publicly.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
Please see the weekness section.
- 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
Weak Reject — could be rejected, dependent on rebuttal (3)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Please see the strengths and weekness section.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed
N/A
- [Post rebuttal] Please justify your decision
N/A
Review #2
- Please describe the contribution of the paper
This paper presents a novel method for reconstructing inner (white matter/grey matter) and outer (grey matter/CSF) cortical surfaces from MRI of fetal brains. The method represents the surfaces as signed distance fields and uses a flow-based method to iteratively deform a surface template to fit the image data. This approach addresses the lower resolution / higher noise images in fetal MRI. It improves mesh regularity and reduces geometric errors compared to two recently published state-of-the-art methods.
- Please list the main strengths of the paper; you should write about 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.
- Tackling a challenging problem that has clinical relevance.
- Provide a novel approach to represent the two cortical surfaces as SDFs, which increases spatial resolution over binary segmentation and reduces surface intersection after deformation.
- Provide a novel loss function that balances accuracy with surface integrity
- Rigorous comparison with significant improvement over existing methods.
- Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
- Pre-training is limited to pre-term babies and newborns because there is limited fetal MRI data. The paper is missing a discussion on how this could impact results.
- The number of iterations limited to 2 due to memory limits. It would be worth doing an ablation study to see if more iterations improve results.
- The value of this work would be increased if data were made available, e.g., through a restricted-access database
- Please rate the clarity and organization of this paper
Excellent
- 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.
- Do you have any additional comments regarding the paper’s reproducibility?
Fetal MRI is not readily available. Understanding the challenges of making such data publicly available, it would still greatly enhance reproducibility if a de-identified version of this data were made available.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
Adding numerical results and a summary of importance of these results to the abstract would strengthen the abstract. As suggested under weaknesses above, future work could include testing the approach with more that two iterations of deformation. The paper is technically motivated (improving surface quality and surfacing speed). It would be helpful to add a clinical motivation. Specifically, why are high quality meshes needed when studying fetal brains? What accuracy is needed? In what time frames does the analysis need to be performed? In the visual comparison in Fig. 3, it is hard to appreciate differences between the three methods. It would be helpful to point out a few specific areas where the results differ and add a discussion on how those differences would impact applications of the proposed segmentation method. Add a justification for why image data and gold standard segmentations are not being released and/or plans for making it available (at least for research and reproducibility).
- 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
Accept — should be accepted, independent of rebuttal (5)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This is a topic that is very relevant to the MICCAI society. The paper provides a new approach for segmenting dual surfaces, i.e., the inner and outer cortical surfaces, where the data is relatively poor (e.g., due to low spatial resolution and noise) and there is significant risk of intersection of the generated surfaces. The evaluation is sound, with results compared to ground truth and two state-of-the-art methods.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed
N/A
- [Post rebuttal] Please justify your decision
N/A
Review #3
- Please describe the contribution of the paper
o This paper proposes a DL-based neonatal surface reconstruction method, utilizing a cycle-consistent, ODE-based pial/white matter refinement network, followed by a 3D-UNet based, age-conditioned deformation and segmentation network. Results show improvements compared to state-of-art ODE-based surface reconstruction methods including CortexODE and CoTAN.
- Please list the main strengths of the paper; you should write about 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 architecture is novel: utilizing one UNet-analogous network and one FCN, the model is capable of generating both segmentations and inner and outer surfaces simultaneously, and the performance reported superior to conventional methods.
- The model exhibits highly robust despite a relatively small number of subjects used in training.
- Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
I really like the idea of cycle-based surface refinement. However, mixing the two models in Fig. 2 (a) and (b) seems redundant. Why is predicting SDF still needed if promising results can be obtained solely from part (a)? From the ablation study, when cycle-consistent model is removed, the model performs comparable, or even better than when it is included, especially considering HD90 and SIF. The authors should address the necessities of combining two models together in a more sounded way. In addition, the predicted field in Fig.2 (b) is not called a ‘SDF’ anymore, but rather a ‘deformation field’.
- Please rate the clarity and organization of this paper
Excellent
- 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.
- Do you have any additional comments regarding the paper’s reproducibility?
None
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review. Pay specific attention to the different assessment criteria for the different paper categories (MIC, CAI, Clinical Translation of Methodology, Health Equity): https://conferences.miccai.org/2024/en/REVIEWER-GUIDELINES.html
None. see my comments above.
- 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
Accept — should be accepted, independent of rebuttal (5)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The structure proposed is novel. However, the cycle-consistent refinement model seems not necessary from the ablation results.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed
N/A
- [Post rebuttal] Please justify your decision
N/A
Author Feedback
N/A
Meta-Review
Meta-review not available, early accepted paper.