Authors

Paula López Diez, Jan Margeta, Khassan Diab, François Patou, Rasmus R. Paulsen

Abstract

The identification of congenital inner ear malformations is a challenging task even for experienced clinicians. In this study, we present the first automated method for classifying congenital inner ear malformations. We generate 3D meshes of the cochlear structure in 364 normative and 107 abnormal anatomies using a segmentation model trained exclusively with normative anatomies. Given the sparsity and natural unbalance of such datasets, we use an unsupervised method for learning a feature representation of the 3D meshes using DeepDiffusion. In this approach, we use the PointNet architecture for the network-based unsupervised feature learning and combine it with the diffusion distance on a feature manifold. This unsupervised approach captures the variability of the different cochlear shapes and generates clusters in the latent space which faithfully represent the variability observed in the data. We report a mean average precision of 0.77 over the seven main pathological subgroups diagnosed by an ENT (Ear, Nose, and Throat) surgeon specialized in congenital inner ear malformations.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43904-9_63

SharedIt: https://rdcu.be/dnwH9

Link to the code repository

https://github.com/paulalopez10/Deep-Diffusion-Unsupervised-Classification-3D-Mesh

Link to the dataset(s)

N/A

Reviews

Review #7

Please describe the contribution of the paper

In this paper, the authors propose an automatic approach for the classification of congenital inner ear malformations. The authors use an unsupervised method to find the latent space representation of cochlear shapes, which allows for their further classification.
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.

In the reviewer´s opinion, the main strength is the uniqueness-novelty of the work. The field adressed by the authors has been vaguely explored. Thus this work is quite useful in the field and can be used as an starting point for future research. The algorithm itself is not novel but the application is.
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 main weakness of the paper is the lack of comparisons, but it is true that is difficult to find works adressing the same issue with public databases to perform such comparisons. The cleariness-writing of the paper could be slightly improved to easy the readness of the work. The authors could provide more details about the values selected for some configuration parameters (more specifically, those related to the shape augmentation process).
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

The author is satisfied with the information provided by the authors.
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/2023/en/REVIEWER-GUIDELINES.html

The main weakness of the paper is the lack of comparisons, but it is true that is difficult to find works adressing the same issue with public databases to perform such comparisons. The cleariness-writing of the paper could be slightly improved to easy the readness of the work. The authors could provide more details about the values selected for some configuration parameters (more specifically, those related to the shape augmentation process).
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?

In the reviewer´s opinion, the main strength is the uniqueness-novelty of the work. The field adressed by the authors has been vaguely explored. Thus this work is quite useful in the field and can be used as an starting point for future research. The algorithm itself is not novel but the application is.
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 presents a novel automated method for classifying congenital inner ear malformations using 3D meshes of the cochlear structure.
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 article is well written and fits well the conference. The novelty of the method is the usage of an unsupervised method for learning a feature representation of the 3D meshes, which captures the variability of the different cochlear shapes and generates clusters in the latent space that faithfully represent the variability observed in the data. Besides, it reachers very promising results in terms of average precision.
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.
- It is not clear how the authors achieved such architecture.
- Comparison with other works from the literature could be improved.
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

This work is reproducible.
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/2023/en/REVIEWER-GUIDELINES.html

This work is very promising, as it achieves competitive results in the detection of congenital inner ear malformations using 3D meshes of the cochlear structure. However, to further validate the efficacy of the approach, additional data could be beneficial in providing a more comprehensive analysis of the detection of each type of malformation. By adding more data, the authors could enhance the study’s ability to identify patterns and classify each malformation accurately, potentially increasing its applicability and usefulness in the clinical setting.
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 usage of an unsupervised learning to detect congenital inner ear malformations is quite interesting and novel. The experiments and methods are well described and shows sufficient. The minor concerns don’t affect the paper quality.
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 #8

Please describe the contribution of the paper

This paper proposed a method for automatically classification of congenital inner ear malformations using solely CT imaging data. The pipeline used DeepDefusion algorithm to learn a latent space representation of the anatonical structure of cochlear to classify different types of inner ear malformations.
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 provides a novel application of DeepDiffusion algorithm in congenital inner ear malformulation classification problem, instead of traditionally malformulation detection or supervised classification 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.
1. There is no algorithmatic novalty in terms of medical imaging computation. All the methods are well-developed.
2. The data used in this paper are highly unbalanced. But this is reasonable as a pilot study of a novel problem.
3. The unsupervised classification might not be appealing to clinicians.
4. The performance evaluation metric is not robust to outliers.
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 method in this study is not complicated and is reproducible. But since the data are not balanced and sample sizes are small for some classes, results might not be robust.
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/2023/en/REVIEWER-GUIDELINES.html
1. To show consistant performance, could you verify this method using publically available dataset?
2. Discuss more on the clinical significance of this study.
3. The last sentance of abstract is not understandable to me, especially ‘compared to the professional diagnosis of an ENT surgeon…’. I assume no human labeled data are involved.
4. Fix minor typo and grammar errors, such as ‘both for the average curve but also for the optimal one for each class’.
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

4
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The results of this study is promising and the application is novel. However, the evidence of model robustness and clnical significance is not adequate for this paper to be accepted.
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

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.

This paper proposed an automatic approach for the classification of congenital inner ear malformations with the mesh structure. The authors used an unsupervised method to find the latent space representation of cochlear shapes. The DeepDefusion algorithm was adopted to learn representation of the anatomical structure of cochlear for classifying different types of inner ear malformations. The reviewers acknowledged its application novelty, study in usually omitted research topics, new 3D data processing application, and interesting experimental results. Almost all reviewers raised questions about its insufficient comparison experiments. However, reviewers also pointed out that it was mainly due to the limited data and research. Finally, the merits outweigh its weakness. Its acceptance may inspire more research and discoveries in 3D medical data processing. The authors are encouraged to release their imaging data and algorithm in their camera ready paper.

Author Feedback

We express our gratitude to the reviewers who have granted us the chance to showcase our manuscript at MICCAI 2023. We sincerely appreciate the dedication and time you invested in evaluating our work, offering valuable suggestions, and providing constructive feedback. Your comments have been thoroughly examined, and we intend to incorporate your insightful suggestions into our work, wherever feasible.

back to top

Unsupervised classification of congenital inner ear malformations using DeepDiffusion for latent space representation