Paper Info Reviews Author Feedback Meta-Review Back to top
List of Papers Browse by Subject Areas Author List

Abstract

Accurate vessel segmentation in Ultra-Wide-Field Scanning Laser Ophthalmoscopy (UWF-SLO) images is crucial for diagnosing retinal diseases. Although recent techniques have shown encouraging outcomes in vessel segmentation, models trained on one medical dataset often underperform on others due to domain shifts. Meanwhile, manually labeling high-resolution UWF-SLO images is an extremely challenging, time-consuming and expensive task. In response, this study introduces a pioneering framework that leverages a patch-based active domain adaptation approach. By actively recommending a few valuable image patches by the devised Cascade Uncertainty-Predominance (CUP) selection strategy for labeling and model-finetuning, our method significantly improves the accuracy of UWF-SLO vessel segmentation across diverse medical centers. In addition, we annotate and construct the first Multi-center UWF-SLO Vessel Segmentation (MU-VS) dataset to promote this topic research, comprising data from multiple institutions. This dataset serves as a valuable resource for cross-center evaluation, verifying the effectiveness and robustness of our approach. Experimental results demonstrate that our approach surpasses existing domain adaptation and active learning methods, considerably reducing the gap between the Upper and Lower bounds with minimal annotations, highlighting our method’s practical clinical value. We will release our dataset and code to facilitate relevant research (https://github.com/whq-xxh/SFADA-UWF-SLO).

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2024/paper/2030_paper.pdf

SharedIt Link: https://rdcu.be/dV5Kc

SpringerLink (DOI): https://doi.org/10.1007/978-3-031-72114-4_8

Supplementary Material: N/A

Link to the Code Repository

https://github.com/whq-xxh/SFADA-UWF-SLO

Link to the Dataset(s)

N/A

BibTex

@InProceedings{Wan_Advancing_MICCAI2024,
        author = { Wang, Hongqiu and Luo, Xiangde and Chen, Wu and Tang, Qingqing and Xin, Mei and Wang, Qiong and Zhu, Lei},
        title = { { Advancing UWF-SLO Vessel Segmentation with Source-Free Active Domain Adaptation and a Novel Multi-Center Dataset } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024},
        year = {2024},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15009},
        month = {October},
        page = {75 -- 85}
}

Reviews

Review #1

  • Please describe the contribution of the paper
    1. The paper considers the vessel segmentation under a novel setting, i.e., images collected from multi-center with domain shift
    2. The paper proposes a new method for advancing cross-center vessel segmentation
    3. The paper also makes contributions to establish and release the dataset used for the experiments
  • 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.
    1. The problem considered in this paper is well motivated and important: the model trained on existing dataset may not perform well on others.
    2. The proposed solution to use source free active domain adaptation is a good fit for the problem
    3. Overall the paper is well written and easy to follow.
    4. Extensive experiments are conducted in the paper to validate the effectiveness of the proposed method
  • 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.
    1. Technical solution is rather simple, especially the selection strategy. It basically looked at the prediction value from the model, ranging from [0,1], and select the ones close to 0.5
  • 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.

  • Do you have any additional comments regarding the paper’s reproducibility?

    N/A

  • 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
    1. Active learning is one of the critical component in the proposed method that select the most valuable patches to label. The paper can consider more analysis on this part, for example, how the model performs w.r.t different C1% and C2%. What’s the overall efforts to label those patches. More visualization on selected patches and analysis on why these patches are hard for source model (e.g., different disease conditions or image quality etc)
    2. The paper replace the patches in the prediction mask Y^t with the corresponding manually labeled patch Y^{Lt} and use it as enhanced pseudo-labels. However, in this setting, the majority of pixels are still from predicted labels and inevitably contains errors. How does the model handle the noisy label and how robust it is.
    3. Figure 3 should also include the model output from [2], which used as source model M^{s} as baseline.
  • 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 Accept — could be accepted, dependent on rebuttal (4)

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    Overall the paper has merits and is of interest to the MICCAI community.

  • 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 #2

  • Please describe the contribution of the paper

    This paper embrace the difficulties of providing accurate estimates of airways geometry. They proposed a novel learning based approach to process airways, compromising point cloud extraction via 2D segmentation and 3D reconstruction from point cloud via neural fields.

  • 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.

    1.investigate the suitability of advanced segmentation techniques for a OCT;

    1. A novel approach to recover 3D geometries from raw point clouds obtained via 2D OCT segmentations; 3 using neural fields to represent 3D geometries from OCT scans
  • 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.
    1. comparative study is not enough, the author mainly compare the segmentation methods, like SAM and nnUnet. No performance comparison with SOTA 3D geometry reconstruction method.
    2. Limited dataset, only 35 OCT scans are evaluated in this paper. Is it possible to extend this method to a larger dataset?
  • 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.

  • Do you have any additional comments regarding the paper’s reproducibility?

    N/A

  • 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
    1. Add experiments to compare SOTA 3D geometry comparison methods.
    2. For section 4.1, confusion from the reason for cylinder coordinates to cartesian coordinate, explanation here.
    3. Please better explain the For figure 4, a) what does the colorful figure mean? What’s the relationship between the colorful one and the plain one? What’s the difference between first raw and second raw?
  • 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 Accept — could be accepted, dependent on rebuttal (4)

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    technical developemtns, paper organization and experiment design.

  • 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 #3

  • Please describe the contribution of the paper

    This paper presents a novel domain adaptation framework and dataset for retinal vessel segmentation in ultra-widefield scanning laser ophthalmoscopy (UWF-SLO) images. In particular, the authors propose a patch-based active domain adaptation approach that uses a Cascade Uncertainty-Predominance (CUP) selection strategy to recommend valuable image patches for labeling and model fine-tuning. The authors also construct a multi-center UWF-SLO vessel segmentation (MU-VS) dataset. Experimental results show that the approach outperforms existing domain adaptation and active learning 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.
    • Relevant dataset for the research community.
    • Novel, straightforward, and effective domain adaptation framework.
    • Very clear and well organized manuscript.
    • Comprehensive experiments and comparisons.
  • 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.

    Overall, I think the paper is a valuable contribution to the field. The only weakness I found is that some details are missing in the Experiments and Results section.

  • 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • Do you have any additional comments regarding the paper’s reproducibility?

    The work is largely reproducible. The source dataset used is publicly available, the authors will release the dataset used for fine-tuning and evaluation, and most of the implementation details are provided. However, some small details are missing in the methodology section.

  • 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

    Overall, I enjoyed reading this paper. It is very well written and organized, with clear contributions and results. I only have two minor comments related to missing details in the methodology section:

    1. The meaning of θs the formula of the loss function Ls(θs, Ωs), in page 4, is never indicated.
    2. In the implementation details, the specific loss function used in the experiments is not mentioned. Please provide this information.
  • 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 paper presents a novel and effective domain adaptation framework for retinal vessel segmentation in ultra-widefield scanning laser ophthalmoscopy (UWF-SLO) images, as well as a relevant multicenter dataset composed of such images. The paper is well written and organized, and the experimental results are comprehensive and convincing.

  • 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

Thank you for the valuable feedback on our manuscript. We appreciate the reviewers’ comments. We will carefully consider them in the final version of our paper to enhance its overall quality.

In particular, we will address the reviewers’ requests for additional details on our loss function and other areas mentioned. By incorporating these modifications, we aim to provide a more comprehensive presentation of our work.



Meta-Review

Meta-review not available, early accepted paper.



back to top