List of Papers Browse by Subject Areas Author List
Abstract
Fibrous cap thickness is a crucial clinical marker for evaluat-
ing carotid plaque vulnerability, emphasizing the importance of accurate segmentation. Intravascular optical coherence tomography (OCT) offers unique capabilities for in vivo visualization of fibrous caps. However, its design for coronary arteries poses challenges in carotid imaging, including larger vessel size, faster blood flow, limited penetration depth,and restricted imaging range, leading to incomplete visualization and degraded image quality from residual blood artifacts. To address these limitations, we propose a dual-coordinate segmentation framework for carotid OCT fibrous cap segmentation. Our approach integrates Cartesian images, which preserve global spatial context and vascular structure, with linear-polar transformed images, which effectively represent the annular geometry of fibrous caps. This dual-coordinate fusion leverages complementary features to mitigate the effects of incomplete vascular walls and blood artifacts, enhancing segmentation accuracy and robustness. To efficiently integrate features from both coordinate systems, we introduce a Cross-Coordinate Feature Fusion Module(CCFFM) that filters and integrates dual-coordinate features, reducing interference from redundant information. Additionally, the Kolmogorov-Arnold Network (KAN) block is incorporated to extract complex nonlinear features while improving the model’s interpretability. Our method achieves state-of-the-art performance on an external carotid OCT dataset, highlighting the potential of OCT for advancing carotid imaging and improving plaque vulnerability assessment.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/3760_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{WanTon_DCKAN_MICCAI2025,
author = { Wan, TongHua and Liu, Sihan and Cai, Yuxin and Chen, Shengcai and Wan, Yan and Hu, Bo and Qiu, Wu},
title = { { DCKAN: A Dual-Coordinate KAN Framework for Fibrous Cap Segmentation on Carotid OCT } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15973},
month = {September},
page = {89 -- 98}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper contributes a new architecture for fibrous cap quantification in Carotid OCT using both cartesian and polar views. They propose a CCFFM module to fuse the views using spatial and channel attention. A KAN layer is used to improve feature representations.
- 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.
- interesting idea to combine multi-view inputs, and the ablation studies clearly show that the additional of their new CCFFM module and inclusion of KAN improve performance when used in conjunction with multi-view inputs.
- well written, clear, and concise throughout.
- quantitative results very good against strong baseline (although concerns about how well optimised these are)
- 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 is based on the claim that using both polar and cartesian views together is superior, but this is not intuitive. Its not clear why one view contains different information to the other, or how they are complementary. is it as simple as they provide features at multiple resolutions? An explanation of this phenomena should be presented to motivate the work.
- another motivating factor is that low penetration depth in OCT means that the modality is weaker for Carotid OCT due to the larger vessel, but its not clear how the proposed work addresses this issue. There should be a clear connection to how the fusion of dual-coordinate features mitigates this issue, with quantitative and qualitative analysis.
- each model trained with the default hyperparameters suggested by the authors of the original work, but these may not be optimal for the dataset used. It’s recommended to do some hyperparameter tuning for each comparative method. Using the same hyperparameters for all experiments is justifiable here.
- No description of internal or external data source, or how ground truth was established.
- 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.
(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?
The architecture is well designed, and the paper well-written but there is a disconnect between the problem statement and the proposed methodology. In the abstract the claim that ‘the fusion of dual-coordinate features mitigates incomplete vascular walls and blood artifacts’ is not supported by evidence.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Reject
- [Post rebuttal] Please justify your final decision from above.
The authors give good explanation as why multi-view is necessary, but I am still not convinced that they have evidence for their central motivation: ‘the fusion of dual-coordinate features mitigates incomplete vascular walls and blood artifacts’
Review #2
- Please describe the contribution of the paper
-
A specialized segmentation architecture tailored for fibrous cap quantification in carotid OCT imaging.
-
A novel cross-coordinate feature fusion module that aggregates multi-scale spatial information from orthogonal coordinate systems, leading to improved boundary delineation.
-
Integration of a Cross-Coordinate Feature Fusion Module (CCFFM) and a Kolmogorov-Arnold Network (KAN) block for carotid imaging adds a layer of technical novelty.
-
- 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.
- Introducing CCFFM for redundancy filtering and KAN for nonlinear feature extraction adds depth and interpretability.
- Accurately segmenting fibrous caps in carotid arteries could significantly improve stroke risk stratification and clinical decision-making.
- 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.
Abstract is text-heavy and could benefit from clearer structure. Although the method achieves enhanced segmentation accuracy and clinical interpretability, it is limited by a lack of dataset diversity and suboptimal computational efficiency for real-time deployment.
- 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?
Novelty could be more explicitly framed against existing work.
- Reviewer confidence
Somewhat confident (2)
- [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.
They addressed rebuttal clearly.
Review #3
- Please describe the contribution of the paper
This paper proposes a KAN framework to address fibrous cap segmentation issue in carotid OCT. A cross-coordinate feature fusion module and KAN network is specifically designed for this study.
- 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 study focuses on segmenting carotid OCT, which is a harder task than conventional coronary segmentation.
- The cross-coordinate feature fusion considers feature representation in both Cartesian coordinate system and polar coordinate system, a specialized design for catheter based OCT system
- The involvement of KAN in image segmentation improves the interpretability of the proposed approach.
- Evaluation is on external carotid OCT.
- The experimental results show outperformances with six SOTA, including nnUNT and UKAN.
- 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.
-(Minor) It will be great to provide more information on data acquisition. For example, which OCT machine is used? What is the resolution? -(Minor) While mean and std are presented in Table 1 and 2, it will be great to provide the p value of a statistical test of the performance between the proposed method and comparison 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 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
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?
The score driving factor is the novel design for catheter-based OCT system
- Reviewer confidence
Very confident (4)
- [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.
Thanks for the thorough rebuttal. I will keep my original score for this high quality paper.
Author Feedback
We thank all reviewers for their insightful comments and suggestions, and have addressed the major concerns as follows: Data Acquisition, Resolution, and Statistical Analysis [R1]: We will include descriptions of our data acquisition protocol, image resolution, and the statistical analysis results in the final version, subject to space permitted. Feature Complementarity in Dual Coordinates [R2]: Cartesian and polar views do not merely correspond to different scales or resolutions; they provide different perspectives, analogous to coronal versus sagittal views of a 3D structure. In the Cartesian view, the vasculature appears as an arc that reflects the true spatial context of the fibrous cap relative to the lipid core and vessel lumen. Conversely, in the polar view, the cap is “unrolled” into a band-shaped structure, highlighting its shape in a flattened geometry. Convolution operation in Cartesian coordinates is inherently translation-invariant, capturing structural shifts across the image, whereas convolution operation in polar coordinates becomes rotation-invariant, capturing structural rotations around the lumen center. These differences in geometric representation and convolutional invariance ensure that Cartesian and polar features capture distinct, yet mutually reinforcing, information about fibrous-cap morphology. Dual-Coordinate Fusion Mitigates Incomplete Vessel-Wall Acquisition [R2]: The polar transformation converts each Cartesian pixel into a radius–angle representation, effectively magnifying regions proximal to the lumen center. This magnification enhances the local detail of the fibrous cap, particularly when portions of its structure are missing or truncated. By amplifying these local features, the network can more accurately delineate the edges and internal boundaries of incomplete caps. Moreover, the complementary nature of Cartesian and polar features is critical to mitigating incomplete vessel‐wall imaging: Cartesian features excel at capturing global shape and macro‐structural context, whereas polar features emphasize local boundary details. Fusing both representations therefore leverages their respective strengths to compensate for missing information. We have quantitatively validated this effect via the ablation study in Table 2. When using only Cartesian inputs, the fibrous‐cap segmentation achieves a Dice coefficient of 0.741; with only polar inputs, the Dice increases modestly to 0.748. By jointly inputting both views and employing our Coordinate‐Complementary Feature Fusion Module (CCFFM), the Dice coefficient rises to 0.779, greatly outperforming either single‐coordinate input. Additionally, we employed neural network interpretability analysis by visualizing network features using Grad‐CAM and LIME to verify the effectiveness of the dual‐coordinate feature fusion. The output feature heatmaps demonstrated that the proposed CCFFM indeed focuses on complementary fibrous‐cap regions in both coordinate systems. Due to space constraints, these qualitative visualizations are not shown here but are available upon request. Hyperparameter Settings [R2]: Prior to submission, we have conducted experiments to evaluate the impact of hyperparameter variations and found that the performance of comparative methods remained stable under modest changes. Therefore, we adopted the recommended settings for all baselines to ensure a fair and reproducible comparison. Ground-truth Annotation [R2]: All annotations were manually delineated by a radiologist with six years of imaging experiences and subsequently reviewed by a senior neurologist with ten years of experiences to ensure label accuracy. Dataset Diversity [R3]: The datasets used for internal and external validation were collected from seven hospitals using two different OCT systems, reflecting great variability and diversity across acquisition conditions. Computational Complexity [R3]: While our model has higher FLOPS and parameter counts, it achieves an inference
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