List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Botond Fazekas, Guilherme Aresta, Dmitrii Lachinov, Sophie Riedl, Julia Mai, Ursula Schmidt-Erfurth, Hrvoje Bogunović
Abstract
Optical coherence tomography (OCT) is a non-invasive 3D modality widely used in ophthalmology for imaging the retina. Achieving automated, anatomically coherent retinal layer segmentation on OCT is important for the detection and monitoring of different retinal diseases, like Age-related Macular Disease (AMD) or Diabetic Retinopathy. However, the majority of state-of-the-art layer segmentation methods are based on purely supervised deep-learning, requiring a large amount of pixel-level annotated data that is expensive and hard to obtain. With this in mind, we introduce a semi-supervised paradigm into the retinal layer segmentation task that makes use of the information present in large-scale unlabeled datasets as well as anatomical priors. In particular, a novel fully differentiable approach is used for converting surface position regression into a pixel-wise structured segmentation, allowing to use both 1D surface and 2D layer representations in a coupled fashion to train the model. In particular, these 2D segmentations are used as anatomical factors that, together with learned style factors, compose disentangled representations used for reconstructing the input image. In parallel, we propose a set of anatomical priors to improve network training when a limited amount of labeled data is available. We demonstrate on the real-world dataset of scans with intermediate and wet-AMD that our method outperforms state-of-the-art when using our full training set, but more importantly largely exceeds state-of-the-art when it is trained with a fraction of the labeled data.
Link to paper
DOI: https://link.springer.com/chapter/10.1007/978-3-031-16452-1_31
SharedIt: https://rdcu.be/cVVpD
Link to the code repository
https://github.com/ABotond/SD-LayerNet
Link to the dataset(s)
N/A
Reviews
Review #2
- Please describe the contribution of the paper
This work introduced a novel SD-LayerNet as a semi-supervised paradigm into the retinal layer segmentation task that makes use of the information present in large-scale unlabeled datasets as well as anatomical priors.
- 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.
Contributions: 1). It is very interesting to propose a semi-supervised paradigm into the retinal layer segmentation task, using of the information present in large-scale unlabeled datasets. 2). In this work, it is promising to notice a variety of reconstruction loss such as Eq (1) and the others proposed in Section 2.2. 3). A comprehensive methodological validation has been included in this work.
- 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.
Major Concerns: 1). Manually tuning parameters The reconstruction loss proposed in this work involves more parameters tuning. The reviewers are afraid that the parameters tuning would influence the performance of proposed method.
2). Further validations In addition, it is more interesting to show the validation of time-consuming of proposed SD-LayerNet with other peer methods.
- Please rate the clarity and organization of this paper
Very Good
- Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance
The reproducibility is good.
The authors provide the link of source code.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html
1). Manually tuning parameters Although the reconstruction loss proposed in Section 2.2 is very promising, it would be challenging to determine all lambda values in reconstruction loss for self-supervising. If all lambda values for reconstruction loss are manually designed, the promising segmentation results of proposed SD-LayerNet should be arbitrary.
2). Further validations In Table.1, the authors provided a methodological validation of proposed method with other two peer methods. These results demonstrate that the performance of proposed methods is better than other peer methods, given the reported segmentation errors and standard deviation. In addition, it is more interesting to show the validation of time-consuming of proposed SD-LayerNet with other peer methods.
- Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making
6
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This paper proposed a novel reconstruction loss that has been proved efficient in the experiments, and a strong experimental validation is provided in this work.
The minor concern is that the proposed reconstruction loss requires more parameter tuning. It would be interesting if the authors could explain or set up more experiments to validate the lambda values.
- Number of papers in your stack
4
- What is the ranking of this paper in your review stack?
1
- Reviewer confidence
Very confident
- [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
The paper proposes a novel medel SD-LayerNet to do semi-supervised retinal layer segmentation in OCT. SD-LayerNet makes use of the information in the large unlabeled datasets as well as anatomical priors. The model use both 1D surface and 2D layer information to train. And this network can also work on nested anatomy and where the thickness of a tissue is measured.
- 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.
This paper solves a very interesting and practical problem, which makes use of unlabeled datasets for semi-supervised learning in retinal layer segmentation task.
This paper is well organized, the idea of innovation is concise and reasonable, and it is easy for readers to follow.
This paper also points out the potential application for other tasks, e.g. inner and outer vessel lumen wall, cardiac wall, knee cartilage, etc.
- 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.
-
Lack of ablation study. If change some of the settings in e.g. anatomical priors, what will happen. The author can do more ablation studies to show the effect of each module of their network.
-
More analysis. Need to show more analysis about the experiment results. The author can discuss more in the results part why this design is better than baseline models.
-
Typos. E.g. Fig 3 caption includes a colored bracket
-
- Please rate the clarity and organization of this paper
Excellent
- Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance
They could reproduce the results with some difficulty.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html
As shown in the weaknesses.
- Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making
6
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
As shown in the weaknesses.
- Number of papers in your stack
5
- What is the ranking of this paper in your review stack?
1
- Reviewer confidence
Confident but not absolutely certain
- [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 #4
- Please describe the contribution of the paper
This paper focuses on the retinal layer segmentation task under the semi-supervised setting. The authors propose Spatial Decomposition Layer Segmentation Network (SD-LayerNet) with two novel contributions: one is a fully differentiable topological engine which facilitates the disentangled representation learning, another is a set of tailored anatomical priors encoded as self-supervised tasks for unlabeled data. The experiments show that the method is able to achieve state-of-the-arts results under the low-data regime.
- 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 authors propose a fully differentiable topological engine, which converts the surface positions to pixel-wise structured segmentations for anatomical representation learning.
- The authors propose a series of self-supervised tasks tailored for retinal layer segmentation based on several anatomical priors.
- The proposed method shows the promise of semi-supervised learning in retina layer segmentation.
- 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 role of the textual factor branch in the anatomy encoder is unclear.
- Too many hyperparameters are involved in the algorithm, including eight loss balance weights and several empirical parameters related to the anatomical priors.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance
good
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2022/en/REVIEWER-GUIDELINES.html
Overall, the paper provides some insightful contributions for semi-supervised retinal layer segmentations. The proposed method provides a possible way for reducing the annotation cost in layer segmentation tasks. The design of the self-supervised tasks in this paper could also inspire future researches in how to incorporate anatomy priors. The main drawbacks lie in that the intuitions behind some design choices are unclear, e.g., the textual factors, and the final algorithm involves too many hypermeters, which may be tricky for tuning. Some mistakes: The notation of the textual factor generation branch seems to be wrong in Figure 1, i.e., conv-t or conv-m?
- Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making
6
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The problem concerned by the paper is meaningful, the proposed method is also technically novel.
- Number of papers in your stack
1
- What is the ranking of this paper in your review stack?
1
- Reviewer confidence
Very confident
- [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
Primary Meta-Review
- Please provide your assessment of this work, taking into account all reviews. Summarize the key strengths and weaknesses of the paper and justify your recommendation. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
The paper proposes Spatial Decomposition Layer Segmentation Network (SD-LayerNet) for semi-supervised retinal layer segmentation. The reviewers were in agreement with the novelty of designing self-supervised tasks with anatomical priors to learning with unlabeled data. Minor concerns were raised regarding the effort needed for manual parameter tuning. Additional ablation studies and analysis were also suggested, , which I believe can be successfully addressed in the final version.
- What is the ranking of this paper in your stack? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers). If this paper is among the bottom 30% of your stack, feel free to use NR (not ranked).
3
Author Feedback
N/A