Authors

Jadie Adams, Shireen Elhabian

Abstract

Statistical shape modeling (SSM) directly from 3D medical images is an underutilized tool for detecting pathology, diagnosing disease, and conducting population-level morphology analysis. Deep learning frameworks have increased the feasibility of adopting SSM in medical practice by reducing the expert-driven manual and computational overhead in traditional SSM workflows. However, translating such frameworks to clinical practice requires calibrated uncertainty measures as neural networks can produce over-confident predictions that cannot be trusted in sensitive clinical decision-making. Existing techniques for predicting shape with aleatoric (data-dependent) uncertainty utilize a Principal Component Analysis (PCA) based shape representation computed in isolation of the model training. This constraint restricts the learning task to solely estimating pre-defined shape descriptors from 3D images and imposes a linear relationship between this shape representation and the output (i.e., shape) space. In this paper, we propose a principled framework based on the variational information bottleneck theory to relax these assumptions while predicting probabilistic shapes of anatomy directly from images without supervised encoding of shape descriptors. Here, the latent representation is learned in the context of the learning task, resulting in a more scalable, flexible model that better captures data non-linearity. Additionally, this model is self-regularized and generalizes better given limited training data. Our experiments demonstrate the proposed method provides an accuracy improvement and better calibrated aleatoric uncertainty estimates than current state-of-the-art models.

Link to paper

DOI: https://link.springer.com/chapter/10.1007/978-3-031-16434-7_46

SharedIt: https://rdcu.be/cVRse

Link to the code repository

https://github.com/jadie1/VIB-DeepSSM

Link to the dataset(s)

N/A

Reviews

Review #1

Please describe the contribution of the paper

A novel framework was proposed for predicting probabilistic anatomical shapes from 3D images based on the variational information bottleneck theory.
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 main strength of this study lies on the novelty of the proposed methodology, which relaxes the linear dependence assumption in existing PCA-based methods. Better calibrated aleatoric uncertainty quantification was achieved without comprising the shape modeling quality. This study is an advance in statistical shape modeling (SSM) and has a great potential to be used as a research tool for SSM analysis.
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.

No major weakness was found.
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

The dataset used for developing and validating the proposed method is not available. The code and model will be made available, and a similar public dataset will be provided for testing the code and model. Results are likely producible.
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
- It would be good to validate the proposed method on an additional dataset, preferably a public dataset.
- Section 3.1 will need further clarification - it is not clear how the ground truth meshes were generated, by who? radiologists or people without training? how many?
- Figure 4 is too small, almost unreadable.
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

7
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The problem this work aims to address is clearly defined - capture the data non-linearity in shape modeling and improve uncertainty quantification. The proposed methodology is novel showing improved performance in terms of both uncertainty quantification and shape modeling compared to the state-of-the-art methods.
Number of papers in your stack

3
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 #3

Please describe the contribution of the paper

This paper proposes to use the variational network to parameterize the statistical shape modeling. It leverages the information bottle neck to enable the learning framework.
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 proposed framework is novel and the difference to the related works are also clearly explained.
2. The validation experiments are well designed.
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. The paper mentions the latent representation is learned, however, it still needs several samples to get the expectations. This part might be over claimed.
2. Some details of the method is not fully exposed.
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 reproducibility of the paper is 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
1. In the method part, it would be more clear that the last layer of the encoder is explained in details.
2. The paper mentions the proposed method is self-regularized. It is not clear to the reviewer why it is self-regularized and this part is not discussed. Also, why the comparison work is not self-regularized is also not discussed. Some discussion can make this more clear.
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

5
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The proposed framework seems to be novel, and the validation is convincing.
Number of papers in your stack

4
What is the ranking of this paper in your review stack?

2
Reviewer confidence

Somewhat 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 #4

Please describe the contribution of the paper

This paper proposes a new method for generating probabilistic shapes from 3D image inputs using a variational information bottleneck (VIB) approach. The claimed advantages are higher accuracy and better uncertainty estimation compared to previous DeepSSM methods. Experiments were performed using a large MRI dataset (1001 scans), with paired ground-truth shape annotations with point correspondence.
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) Thorough explanations for proposed vs previous methods (2) Good shape-predictive performance and uncertainty estimation of the proposed method (3) Simple architecture / Loss (Conv/MLP and VAE ELBO loss with a \beta term for KL divergence)
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 seem to be two main differences for VIB-DeepSSM vs other methods - (a) loss and (b) MLP decoder. It’s unclear which is contributing more significantly to the improved shape prediction and uncertainty estimation. (2) Unclear whether the dataset is public. If not, may be worthwhile to include more info about acquired images. (3) No mention of Figure 4 in the main text - the x-axis seems to be Training size, but PPCA-Offset-DeepSSM model doesn’t change values at all throughout different training sizes. Why is this the case?
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

Reproducibility seems 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

(1) To address weakness #1, it may have been a good idea to perform ablation studies for the two different contributions separately. (2) It would be good to clarify whether the dataset is public/private and to include more descriptions if it is private. (3) It would be good to explain figure 4 (i.e. shape accuracy) a bit in the main text.
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

5
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

Overall, the paper was well-written and included relevant explanations and experiments that demonstrated the uniqueness and advantages of the proposed method, respectively. I have some reservations about whether or not the proposed method is novel enough to be considered a new method (conv + MLP and beta-VAE loss), but I think in conjunction with the shape modeling task and performed experiments, it may still be of interest to the MICCAI community.
Number of papers in your stack

7
What is the ranking of this paper in your review stack?

3
Reviewer confidence

Somewhat 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 overall comment of this paper is positive. All reviewers agreed that the proposed approach to generating probabilistic shapes from 3D images based on variational information bottleneck theory is new and interesting. Experimental results were convincing and well demonstrated that this method was superior to the-state-of-the-art. The authors are strongly encouraged to incorporate all reviewers’ constructive feedback carefully into a revised 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

back to top

From Images to Probabilistic Anatomical Shapes: A Deep Variational Bottleneck Approach