List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Jianpeng Zhang, Xianghua Ye, Jianfeng Zhang, Yuxing Tang, Minfeng Xu, Jianfei Guo, Xin Chen, Zaiyi Liu, Jingren Zhou, Le Lu, Ling Zhang
Abstract
Lung cancer is a leading cause of death worldwide and early screening is critical for improving survival outcomes. In clinical practice, the contextual structure of nodules and the accumulated experience of radiologists are the two core elements related to the accuracy of identification of benign and malignant nodules. Contextual information provides comprehensive information about nodules such as location, shape, and peripheral vessels, and experienced radiologists can search for clues from previous cases as a reference to enrich the basis of decision-making. In this paper, we propose a radiologist-inspired method to simulate the diagnostic process of radiologists, which is composed of context parsing and prototype recalling modules. The context parsing module first segments the context structure of nodules and then aggregates contextual information for a more comprehensive understanding of the nodule. The prototype recalling module utilizes prototype-based learning to condense previously learned cases as prototypes for comparative analysis, which is updated online in a momentum way during training. Building on the two modules, our method leverages both the intrinsic characteristics of the nodules and the external knowledge accumulated from other nodules to achieve a sound diagnosis. To meet the needs of both low-dose and noncontrast screening, we collect a large-scale dataset of 12,852 and 4,029 nodules from low-dose and noncontrast CTs respectively, each with pathology- or follow-up-confirmed labels. Experiments on several datasets demonstrate that our method achieves advanced screening performance on both low-dose and noncontrast scenarios.
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43904-9_20
SharedIt: https://rdcu.be/dnwGY
Link to the code repository
N/A
Link to the dataset(s)
N/A
Reviews
Review #2
- Please describe the contribution of the paper
This paper proposes a method to leverage contextual parsing and prototype-based learning to classify malignant and benign nodules. The contextual parsing segments lung structures to provide rich information. Prototype-based learning recalls references for comparison analysis. Two data cohorts are used for model development and evaluation.
- 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 is using a novel collaborative model towards parsing lung structures and predict nodule malignancy. The organization is good overall. The combination of parse and recall module helps improve nodule classification performance.
- 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.LDCT and NCCT are not well defined in this paper, which have intersection CT scope. Please confirm whether NLST is contrast CT or NOT. (see Section 3.1: “Unlike NLST, this dataset is non-contrast chest CT”)
2.Methods evaluated in this study should be considered to allow more comprehensive analysis of the results [1] (e.g., Accuracy, Sensitivity, Specificity and Precision). Only AUC was assessed here. [1] Yu, Marco, et al. “Reporting on deep learning algorithms in health care.” The Lancet Digital Health 1.7 (2019): e328-e329.
3.This paper do not show the segmentation performance of trachea (especially small airways) generated by the TotalSegmentator in Table and Figure.
4.It is not clear if CT resampling has been performed before cropping the input size of 32x48x48. NLST data was collected two decades ago with large slice-thickness (poor resolution).
- How to deal with class-imbalance (benign vs malignant) when cross-entropy loss is used for classification?
- 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 seems OK. It would also be valuable if they made the dataset with annotations available.
- 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.This paper should present more comprehensive evaluation metrics for statistical results. 2.Vessel segmentation can be classified into artery and vein trees to provide more detailed information for nodule malignancy prediction. 3.Testing the performance of trachea segmentation generated by TotalSegmentator. 4.Reference 16 and 17 are same.
- 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?
Please see the detailed comments in above sections.
- 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 #1
- Please describe the contribution of the paper
This paper proposes the PARE model, which aims to accurately predict whether a lung nodule is malignant by analyzing contextual information from the nodule and recalling previous diagnostic knowledge. The model achieves this by parsing the contextual information from the nodule itself and recalling the previous diagnostic knowledge to explore related benign or malignancy clues. Additionally, the paper curates a large-scale pathological-confirmed dataset with up to 13,000 nodules to fulfill the needs of both LDCT and NCCT screening scenarios. The model achieves outstanding performance in predicting malignancy in both scenarios and demonstrates strong generalization ability in external validation. The proposed method surpasses other advanced methods and achieves state-of-the-art results.
- 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 PARE model accurately predicts whether a lung nodule is malignant by analyzing contextual information from the nodule and recalling previous diagnostic knowledge. The method consists of two modules: context parsing and prototype recalling. The context parsing module segments the context structure of nodules and aggregates contextual information for a more comprehensive understanding of the nodule. The prototype recalling module utilizes prototype-based learning to condense previously learned cases as prototypes for comparative analysis. The overall idea is novel and the implementation is practical for clinical application.
- A large-scale dataset is collected.
- Extensive experiments are implemented to demonstrate the advantages of the proposed methods.
- The paper is well-written and very easy to understand
- 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 paper proposes a framework that utilizes cross-prototype attention. Although this idea is novel, there is a concern about separating the training of UNet and the following steps. More details are mentioned at section 8.
- 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
Enough details are provided for reproducing the paper.
- 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 author proposes a framework that combines contextual semantic information and uses the Vit structure to globally learn the connections between various patches, resulting in final results using a query method. The idea is innovative. However, there is a concern regarding the base of the entire framework, which lies in the quality of segmentation network. The advantages of traditional segmentation methods is to use benign and malignant tumor masks for training, as it learns the differences between benign and malignant tumors based on segmentation, especially the details. In the proposed framework, the segmentation network is used to segment tumors and other organs. The training of Unet with the following steps is trained separately. Please correct me if I am wrong. Have you considered an end-to-end training model? What are the concerns if it has not been implemented?
- 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?
- Idea is novel. The PARE model predicts whether a lung nodule is malignant by analyzing contextual information from the nodule and recalling previous diagnostic knowledge.
- Extensive experiment shows state-of-the-art performance in predicting malignancy and surpasses other methods.
- The paper collects a large-scale pathological-confirmed dataset with up to 13,000 nodules to fulfill the needs of both LDCT and NCCT screening scenarios. The paper is well-written and easy to understand, but there is minor concern about separating the training of UNet and the following steps and the quality of the segmentation network in the proposed framework.
- 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 presents a PARE model for lung nodule malignancy prediction. It essentially includes a UNet based context parsing module for parsing the contextual information from the nodule itself, and a prototype learning based recalling module for recalling the previous diagnostic knowledge for comparative analysis. Authors validate the method on LDCT and NCCT datasets to meet the needs of both low dose and non-contrast screening.
- 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 the proposed method is that a prototype learning based recalling module is introduced for recalling the previous diagnostic knowledge to explore potential inter-level clues as an additional discriminant criterion for the new case.
- 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 proposed method uses UNet to perform lung nodule segmentation but the network structure and segmentation procedure are not introduced. The center of each cluster is used as prototype but the effectiveness of the cluster results is not discussed. Only AUC is used as indicator to evaluate the performance, other indicators are recommended to supplement for a comprehensive evaluation of performance.
- Please rate the clarity and organization of this paper
Satisfactory
- 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 experimental settings and some parameters are not defined.
- 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
For future work, I would recommend: 1) provide a more detailed description of the method including the network structure and parameters. 2) provide more indicators for a comprehensive evaluation of performance. 3) discuss the effectiveness of the cluster results used as prototype.
- 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?
Limited explanation of method details.
- 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 studying a new concept which is a method to simulate the diagnostic process of radiologists in an application of distinction between benign and malignant lung nodules in thoracic CT.
Author Feedback
We appreciate the positive feedback and suggestions from both the AC and reviewers on our work, and we will consider their suggestions to further improve the quality of this work in either the final version or future work.