List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Tianyu Zhang, Luyi Han, Anna D’Angelo, Xin Wang, Yuan Gao, Chunyao Lu, Jonas Teuwen, Regina Beets-Tan, Tao Tan, Ritse Mann
Abstract
Magnetic resonance imaging (MRI) is the most sensitive technique for breast cancer detection among current clinical imaging modalities. Contrast-enhanced MRI (CE-MRI) provides superior differentiation between tumors and invaded healthy tissue, and has become an indispensable technique in the detection and evaluation of cancer. However, the use of gadolinium-based contrast agents (GBCA) to obtain CE-MRI may be associated with nephrogenic systemic fibrosis and may lead to bioaccumulation in the brain, posing a potential risk to human health. Moreover, and likely more important, the use of gadolinium-based contrast agents requires the cannulation of a vein, and the injection of the contrast media which is cumbersome and places a burden on the patient. To reduce the use of contrast agents, diffusion-weighted imaging (DWI) is emerging as a key imaging technique, although currently usually complementing breast CE-MRI. In this study, we develop a multi-sequence fusion network to synthesize CE-MRI based on T1-weighted MRI and DWIs. DWIs with different b-values are fused to efficiently utilize the difference features of DWIs. Rather than proposing a pure data-driven approach, we invent a multi-sequence attention module to obtain refined feature maps, and leverage hierarchical representation information fused at different scales while utilizing the contributions from different sequences from a model-driven approach by introducing the weighted difference module. The results show that the multi-b-value DWI-based fusion model can potentially be used to synthesize CE-MRI, thus theoretically reducing or avoiding the use of GBCA, thereby minimizing the burden to patients. Our code is available at
https://github.com/Netherlands-Cancer-Institute/CE-MRI.
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43990-2_8
SharedIt: https://rdcu.be/dnwLi
Link to the code repository
https://github.com/Netherlands-Cancer-Institute/CE-MRI
Link to the dataset(s)
N/A
Reviews
Review #2
- Please describe the contribution of the paper
In this paper, the authors propose a novel processing method named hierarchical fusion module that synthesizes Breast Cancer Enhanced MRI images based on the DWI technique, with the advantage of avoiding certain contrast agents
- 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 paper is the novel processing method named hierarchical fusion module that overcomes other proposed methods according to the selected metrics
- 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 is the Ground Truth defined (a Contrast Enhanced reference image). A synthetic CE-MRI set is mentioned in the results sections that were used for comparison, but it lacks information about this set. Is this the one used as GT? How was obtained? how accurate is it?
- 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
As mentioned, the reference synthetic CE-MRI used for comparing results is not explained, to reproduce the experiment.
- 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 results section both tables indicate that the method outperforms previous results however figure 3 should confirm this. The image description is poor and is not visually clear for the reader the advantages of the method. For example, there are red circles in the figure, what do they mean? do they indicate the method performs better? I should recommend a detailed description of quality performance on the same image description
- 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 main contribution of this paper is the novelty processing method (hierarchical fusion module), and there are some missing points (like the description of quality results in the figure and description of the synthetic data), however the contribution is well described, and the missing points just need an extra explanation
- 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 multi-sequence fusion network to synthesize CE-MRI based on T1-weighted and diffusion weighted MR imaging. The proposed method is evaluated on a local dataset containing 765 patients with breast cancer.
- 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.
Compared to previous methods of synthetic CE-MRI the proposed method uses DWI, which may be valuable if only selected sequences from MRI is available. DWI may provide useful information on tissue microstructure and may provide alternate information not in conventional sequences valuable to synthesize CE-MRI.
- 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 overall idea is interesting. However, the added effectiveness and performance over the different existing methods is not that high. Additionally, it is not clear how often “multi-b value” MRI performed in routine breast MR imaging?
- This was single institution study, limiting the generalizability. Additionally, the study included invasive breast cancers only.
- 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
No code or data was provided along with 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 study proposes to use multi-b-value DWI to synthesize CE-MRI for the first time. However the described methods Chung et al. and MMgSN-Net also include DWI. Recommend to clarify the differences of the b values used.
- In table 1 and table 2, the results should be specified as mean +/- SD
- Figure 3, some of the images are too small to see the lesion adequately and could be better cropped
- 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 work addresses an interesting problem and the method has some novelty.
- 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
The authors propose a generative model to generate the DCE-MRI from T1WI and multi-b-value DWI. The achievements of this study may be of good clinical value by allowing patients to avoid the adverse effects of contrast agents.
- 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 generative model to generate the DCE-MRI from T1WI and multi-b-value DWI. The achievements of this study may be of good clinical value by allowing patients to avoid the adverse effects of contrast agents.
- 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.
- Many small grammatical problems appear which brings some confusing sentences.
- Some of the terminology in the description is not accurate.
- DCE-MRI usually has images of multiple time points, representing different stages of contrast inflow and outflow from the lesion. In this manuscript, the DCE-MRI image was generated for only a one-time point. The question of which time point the DCE-MRI image was generated has not been described. In addition, is there any relevant support for the selection of DCE-MRI at that time point? Why is it not possible to generate MRI images for all time points?
- 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 proposed model used T1WI and multi-b-value DWI images as input, while the T1WI information is not reflected in the title.
- The term “contrast-enhanced MRI (CE-MRI)” is less common and we usually use DCE-MRI.
- Some sentences in this manuscript are not precisely expressed, such as “However, the use of gadolinium-based contrast agents (GBCA) requires iv-cannulation, which is a burden to patients, time consuming and cumbersome in a screening situation” or “We invented (a or the) hierarchical fusion module,”
- You said, On the other hand, the information provided by multi-sequences may be redundant and may not contain the relevant information of CE-MRI. But is there a basis for the premise of this hypothesis?
- Another confusing sentence: invented hierarchical fusion module, weighted difference module and multi-sequence attention module to enhance the fusion at different scale, to control the contribution of different sequence and maximising the usage of the information within and across sequences. Lots of little mistakes, like plural nouns without the ‘s’.
- What’s the meaning of “ln function”? Is it the log function?
- Why use a generative adversarial training strategy? A direct calculation of the MSE loss between DCE-MRI and the generated image should be sufficient to generate the desired image.
- The loss function LD(G, D), used for training the Discriminator, should have an extra negative sign.
- In the sentence “The numbers of filters are 32, 64, 128, 256 and 512, respectively”, the numbers should be revised into channels.
- Can’t understand the part of “2.3 Visualization”.
- In the sentence “The trade-off parameter λ1 was set to 100 during training, and the trade-off parameter of the reconstruction loss in the reconstruction module is set to 5.”, what’s the difference between the two trade-off parameters?
- “The batch was set to 8 for 100 epochs”? Do you want to say “the batch size”？
- What is the meaning of “thereby optimizing logistics and minimizing potential risks to patients.”?
- In the right part of Fig.3., you note the (ceT1-T1). What is the meaning of (ceT1-T1)? Is it the subtracted image between the GT and the model outputs? If they are the subtracted images, the highlighted parts of them mean more difference. If they are the output images of models, the details inside the chest are worse.
- Finally, DCE-MRI usually has images at multiple time points, representing different stages of contrast inflow and outflow from the lesion. In this manuscript, only one DCE-MRI image at the one-time point was generated. So, at what point in time did you generate the image? Whether there is a basis for relevant support. Why can’t all time points be generated?
- 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
- Please check and optimize the grammar and terminology aspects, which will improve the reading experience for reviewers and readers.
- Constructing more solid theoretical hypotheses related to DCE-MRI images at multiple time points is the cornerstone of this study.
- In the process of applying some modules or techniques, the advantages of them need to be more clearly explained, such as the generating adversarial strategy.
- 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 mainly consider the following factors:
- the value or significance of the research objectives.
- Innovation in methodology and data, etc.
- Reliability of the hypothesis underlying it.
- The strength and reliability of the experimental results.
- The level or readability of the writing.
- 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.
The paper presents a hierarchical fusion module to synthesize post-contrast T1W MRI images from a combination of pre-contrast T1W and multi-b-value DWI data. The novelty resides in both the combination of pre-contrast T1W and DWI data and the valuable clinical application that can reduce the need for contrast material. All reviewers agreed on the high quality of the paper while raising concerns about the definition of the reference to compare along with the clarity of the visual improvement achieved by the proposed approach. Authors should address these issues in the final submission.
Author Feedback
We greatly appreciate the area chairs and reviewers for their efforts and comments. We are encouraged that reviewers found the novelty of our model, as well as the importance and good clinical value of our work. We will try our best to correct or clarify all the concerns of the reviewers in the final version to improve the quality of this article. Some specific issues are listed below.
Q1. How is the Ground Truth (GT) defined? How to interpret the visualization results? A1. Contrast-enhanced images were selected as GT. In the visualization, the T1-weighted images and the contrast-enhanced images were subtracted to obtain a difference MRI to clearly reveal the enhanced regions in the CE-MRI. As shown in Figure 3, the red circle represents the highlighted area, which proves that our method can effectively synthesize contrast-enhanced images, highlighting the same parts as GT, while other contrast methods fail to do so.
Q2. Clarify differences in b-values used in comparative methods. Table 1, table 2 and figure 3. A2. Although the method described by Chung et al. also includes DWI, they only use DWI images with two different b-values (b=0 and another specified b-value). Also, they did not specifically design models or modules to improve efficiency. In contrast, our study used multi-b-value DWI images (b=0, 150, 800, 1500) for the first time, and innovatively designed the Hierarchical fusion generation module, weighted difference module, and multi-sequence attention module, which effectively improved the synthesis efficiency. The results in table 1 and 2 were specified as mean +/- SD. We will try to crop some of the images in Figure 3 to make them clearer.
Q3. Multiple time points of DCE-MRI images. A3. Yes, DCE-MRI usually has images of multiple time points, representing different stages of contrast inflow and outflow from the lesion. In this study, our goal was to synthesize contrast-enhanced images and then subtract T1-weighted and contrast-enhanced images to obtain differential MRI to clearly show enhanced regions in CE-MRI. Only one time point (the first time point), the ultrafast image, is selected as GT, which is sufficient for our purpose. Follow-up work will expand the research content, such as synthesizing dynamic contrast-enhanced MRI at multiple time points. Thank you for your suggestion.