Authors

Shangxuan Li, Chichi Li, Yu Du, Li Ye, Yanshu Fang, Cheng Wang, Wu Zhou

Abstract

The recognition of dental pulp calcification has important value for oral clinic, which determines the subsequent treatment decision. However, the recognition of dental pulp calcification is remarkably difficult in clinical practice due to its atypical morphological characteristics. In addition, pulp calcification is also difficult to be visualized in high-resolution CBCT due to its small area and weak contrast. In this work, we proposed a new method of tooth segmentation, identification and pulp calcification recognition based on Transformer to achieve accurate recognition of pulp calcification in high-resolution CBCT images. First, in order to realize that the network can handle extremely high-resolution CBCT, we proposed a coarse-to-fine method to segment the tooth instance in the down-scaled low-resolution CBCT image, and then back to the high-resolution CBCT image to intercept the region of the tooth as the input for the fine segmentation, identification and pulp calcification recognition. Then, in order to enhance the weak distinction between normal teeth and calcified teeth, we proposed tooth instance correlation and triple loss to improve the recognition performance of calcification. Finally, we built a multi-task learning architecture based on Transformer to realize the tooth segmentation, identification and calcification recognition for mutual promotion between tasks. The clinical data verified the effectiveness of the proposed method for the recognition of pulp calcification in high-resolution CBCT for digital dentistry.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43904-9_68

SharedIt: https://rdcu.be/dnwIg

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #1

Please describe the contribution of the paper

The authors have integrated tooth segmentation and recognition of dental pulp calcification in high-resolution CBCT images in an end-to-end manner using a multi-task Transformer network, which first down scales, segments and uses it to recognize the tooth region as the input for fine segmentation to reduce the required computational power. They have proposed tooth instance correlation and triple loss for better recognition of calcifications. The segmentation part is validated by comparing to some conventional 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.
- Novel approach for detecting pulp calcification using deep learning by benefitting from collaborative labels from experienced dentists that can be beneficial for less-experienced ones.
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 body of the paper has exceeded the 8-page limit by two lines.
- weak validation for pulp detection section: There is only an ablation study showing how different modules of the network affect the output.
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 authors have claimed that they will share their github repo.
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
- Although there is not a deep learning-based study in the literature to compare the results with, authors could validate their method’s result through a new group of experts.
- It would be beneficial if the figures had longer and self-contained captions; e.g. Figure 2. and 3.
- There are some minor issues that can be resolved by a general edit; e.g. CBCT is not defined in the abstract, ICT is defined twice, some grammatical errors, etc.
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?

I believe the results show the merits of the method, but there is a need for more evaluation.
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 #2

Please describe the contribution of the paper

The authors proposed a deep learning pipeline for the automatic detection of dental pulp calcification.
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 paper integrated a segmentation network and a classification network to form a pipeline that can detect dental pulp calcification from CBCT in an end-to-end manner.
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 is poorly written, and it raises concerns about whether individuals who do not speak the authors’ native language would be able to understand it. It is premature for the paper to be published at this time.
2. Based on my understanding of the paper, the authors used a two-step segmentation approach. However, it is unclear from Fig. 2 where the two steps are. Is the high-resolution CBCT cropped based on the segmented results shown in Fig. 2 and then fed into another fine-grained segmentation network?
Please rate the clarity and organization of this paper

Poor
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

Assuming that the authors follow through with their announcement in the paper and share the code, it should be feasible to reproduce the results.
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 authors could potentially utilize tools such as Grammarly or even chatGPT to enhance their writing. Although there may not be too many grammatical errors, a considerable number of words have been misused.
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

3
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

A considerable number of words are misused in the paper. It is premature for the paper to be published at this time.
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

This paper proposes a coarse-to-fine tooth segmentation and identification method for high-resolution CBCT data, based on the transformer structure, as well as a method for pulp calcification identification. The experimental results show certain effectiveness.
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.

A coarse-to-fine processing method is proposed in this paper, which can segment and identify calcification regions simultaneously on high-resolution CBCT.
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.

Small dataset size may lead to generalization issues.
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 authors reported most of the experimental parameters, thus making the study highly 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

In the experimental section, the final results are not sufficiently compared with other methods. It is recommended to compare with more relevant research, especially for the classification results of calcified points. It is suggested to collect more data as the dataset used in this study is relatively small, which may lead to potential generalization issues. The use of transformer structures can be explored to collect data from different modalities and apply them to this task.
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 method achieved good results and has certain innovation.
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.

This is a good paper developing a new method for tooth segmentation and identification, and pulp calcification recognition by using a transformer model in high-resolution dental corn-beam CT images.

Author Feedback

N/A

back to top

Transformer-based tooth segmentation, identification and pulp calcification recognition in CBCT