Back to top List of papers List of papers - by topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Christina Bukas, Bailiang Jian, Luis Francisco Rodríguez Venegas, Francesca De Benetti, Sebastian Rühling, Anjany Sekuboyina, Jens Gempt, Jan Stefan Kirschke, Marie Piraud, Johannes Oberreuter, Nassir Navab, Thomas Wendler
Abstract
Symptomatic spinal vertebral compression fractures are often treated by osteoplasty where a cement-like material is injected into the bone to stabilize the fracture, restore the vertebral body height and alleviate pain. Leakage is a common complication and may occur due to too much cement being injected. Here, we propose an automated patient-specific framework that can allow physicians to calculate an upper bound of the volume of cement for particular types of VCFs and estimate the optimal outcome of osteoplasty. The framework uses the patient CT scan and the segmentation label of the fractured vertebra to build a virtual healthy spine. Firstly, the fractured spine is segmented with a three-step Convolutional Neural Network architecture. Next, a per-vertebra rigid registration to a healthy reference spine restores its curvature. Finally, a GAN-based inpainting approach replaces the fractured vertebra with an estimation of its original shape, the volume of which we use as an estimate of the original healthy vertebra volume. As a clinical application, we derive an upper bound on the amount of bone cement for the injection. We evaluate our framework by comparing the virtual vertebrae volumes of ten patients to their healthy equivalent and report an error of 3.88±7.63%. The presented pipeline offers a first approach to a personalized automatic high-level framework for planning osteoplasty procedures.
Link to paper
DOI: https://doi.org/10.1007/978-3-030-87202-1_51
SharedIt: https://rdcu.be/cyhQ6
Link to the code repository
https://github.com/christinab12/bone-cement-injection-planning
Link to the dataset(s)
N/A
Reviews
Review #1
- Please describe the contribution of the paper
This paper presents a method, based on CNN, to visualize what should be a the healthy shape of a fractured vertebra from a CT-Scan. This allows to compute the maximum quantity of cement to inject in the bone.
- 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 application is interesting. Results are encouraging.
- 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 context is too quickly presented.The methodology is not detailed. We do not see any clinical results.
- 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
It is written at the beginning of section 2 that “code for running the pipeline will become publicly available upon acceptance”.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
- Introduction
The context is too quickly presented.
- Technical concepts are not explained or in an ambiguous way: what is “virtual vertebra reconstruction” (reconstruction is used for many things in image processing), define “straightening” and “inpainting”.
- Data are not detailed. What can we see exactly in a CT image (bone structure, fracture)? Some surface visualization of vertebrae and spine will help to understand the concept of “straightening” in 3D. What is the “fractured vertebra label”?
- Methodology
The methodology is not detailed.
- Why did you select “the approach of Sekuboyina”?
- You write that you “reimplement the approach” but when you describe it (very vaguely), you explain that you use a “modified version of the BtrflyNet” without giving any detail.
- You never define the objective of “Spine Straightening”. You cite 2 references but they focus on registration-based segmentation.
- We “follow a similar approach”. Which approach?
- Why using the label of fracture as an input parameter?
- The sentence “We calculate the distance (between what?) using the first two principal components of the centroïd’s coordinates (did you perform PCA?)” is incomprehensible.
- How do you perform automatic rigid registration between the vertebrae and the segmentation masks?
- Explain that Algorithm 1 performs a kind of interpolation of the displacements of the bony structures and apply them to soft tissues.
- The sentence “we tackle the inpainting as a multi-task learning problem (which tasks?), and train a model (so, it is machine-learning based method that you are going to use?)” is incomprehensible.
- Instead of explaining the architecture of Yu et al. with words, you should present a Figure. It would allow you to emphasize your own contribution sas the Dice loss.
- The sentence “while for the segmentation (it was in the first step that we perform segmentation?), the softmax outputs (without any general presentation of the architecture, we cannot know where are the softmax functions) of the segmentation layer are averaged before acquiring the final predictions (what are these predictions?)” is incomprehensible.
- Experimental Setup
- We can find a lot of details in this section.
- Results
- Figure 3 is interesting but a 3D surface rendering of the vertebrae would be better to understand the straightening step in 3D.
- Figure 4 is poor. You should display several slices and focus more on the fractured vertebra.
- Explain how and why you compute SSIM and PSNR? What is the reference image?
- There is no clinical assessment of the result: is you error small enough to use your pipeline for defining the maximum cement amount?
- Discussion
- This section is not interesting as this neither a discussion (no reference), no analysis of results, nor a presentation of future work or perspectives.
- Please state your overall opinion of the paper
reject (3)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The context is too quickly presented.The methodology is not detailed. We do not see any clinical results.
- What is the ranking of this paper in your review stack?
5
- Number of papers in your stack
5
- Reviewer confidence
Confident but not absolutely certain
Review #2
- Please describe the contribution of the paper
The paper proposes a method for predicting the vertebral shape after a preferable osteoplasty, which would help planning a preferable amount of cement injection.
- 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 proposed framework for straightening and reconstructing (inpainting) the damaged vertebrae is quite novel and proposes a good solution for the problem that is closely related to a real clinical issue (i.e., surgical planning). The evaluation for each step was done thoroughly and demonstrated promising results.
- 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 experiment did not collect any information about the injected cement, most importantly the amount of cement that was actually injected during the osteoplasty. Also, the severity of osteoporosis changes the amount of cement need to be injected. Thus, the preferable amount of cement is not determined solely from the predicted healthy vertebral shape. I understand this is out of scope of this MICCAI submission, but there are some overstatements in the abstract and introduction which may discourage the reader.
- 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 reviewer thinks there is no problem with the checklist.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
This is a nice work addressing the highly clinically relevant problem using deep neural networks. I have three comments. 1) Due to the reason that I mentioned in 4 (weakness), I think some of the authors statement about the cement volume estimation (e.g., third sentence from the end in abstract, “we estimate the maximum amount of bone cement for injection”) is overstatement. I think its better to reword all these sentences except for the part referring to the potential application (which I totally agree). 2) I was confused with the notation “pre-fractured” (in section 4, 5, Fig. 3, 5, etc.), which reminded me of a patient before the fracture (i.e., a patient when they were healthy). I think what the authors meant here is “pre-osteoplasty” or “pre-op” (or maybe “fractured”), etc. 3) There was almost no explanation about the “spine atlas” that the author used in straightening step. Is this an open data set? How was it constructed?
- Please state your overall opinion of the paper
strong accept (9)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The proposed framework for a planning of the complex spine surgery is quite nice and has a wide range of potential applications not only for the osteoplasty. After removing some overstatements and adding an explanation about the spine atlas (which I mentioned above), I think this paper is worth publication.
- What is the ranking of this paper in your review stack?
1
- Number of papers in your stack
3
- Reviewer confidence
Very confident
Review #3
- Please describe the contribution of the paper
This work proposes an automated patient-speci c framework that can allow physicians to calculate an upper bound of cement for the injection and estimate the optimal outcome of osteoplasty.
- 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 idea is reasonable with high clinical value.
- 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.
- Patients with VCFs may have abnormal spinal alignments, such as kyphosis. Virtual Spine Straightening only raised the height of the fractured vertebra without reconstructing the normal shape of the spine.
- It is difficult to reconstruct a fractured vertebra into a completely normal one. It is necessary to comprehensively evaluate the severity of the fracture, the duration of the disease, and the bone mass etc. Therefore, it is easy to cause bone cement leakage according to the method in this article.
- 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
It’s an ok 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://miccai2021.org/en/REVIEWER-GUIDELINES.html
- Is it necessary to segment and reconstruct the posterior column of the fractured vertebra?
- How did the author obtain the CT scan of pre-fractured? Is this ethical?
- The number of subjects is limited. Please increase the sample size.
- Please state your overall opinion of the paper
borderline accept (6)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Substantial revision is needed for the current version.
- What is the ranking of this paper in your review stack?
3
- Number of papers in your stack
5
- Reviewer confidence
Confident but not absolutely certain
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.
Assessing the amount of cement to be injected during osteoplasty is an important clinical question that the authors tackled with an exciting approach. However, the authors might overstate the capacity of the presented approach, because the maximal amount of cement injected into a fractured vertebra will not be defined only by the best fitted healthy vertebra. Nevertheless, I see this method as an important milestone towards the automatic estimation of the cement volume. Moreover, inpainting a fractured vertebra with a patient-specific healthy vertebra might also go beyond osteoplasty. Since inpainting is a crucial step in the approach, the authors needed to put more effort into comparison to SOTA methods. The manuscript could also benefit from rewriting and redrawing the figures, as pointed out by reviewers. Therefore, I would strongly recommend the authors to include as many reviewers’ comments as possible into the camera-ready version of the manuscript. Finally, the authors should also introduce a brief statement about the limitations of the method, as well as future steps towards identifying not only the upper bound of injected cement but also its optimal amount.
- 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).
1
Author Feedback
We thank the reviewers for giving us the opportunity to present our manuscript to MICCAI 2021. We are grateful for the time and effort you spent on carefully assessing our manuscript and providing insightful suggestions and constructive feedback. We have studied your comments and plan to adapt our work accordingly to incorporate your suggestions wherever possible. More specifically, we would like to point out the following responses and clarifications:
Comments to Metareviewer #2 (1) We thank you for noting that the capacity of our approach may be overstated; this is a crucial point. We will adjust the phrasing in the discussion to more carefully state the range of validity of our results. To obtain a prediction of the actual amount to be inserted, we will need more data with complications (i.e. due to too little cement injected or leakage). Increasing the specificity of our result will be part of future work.
(2) We will include in our introduction other applications for the vertebra inpainting. (3) We will extend our manuscript to outline existing SOTA for inpainting and explain why our method is superior to other approaches for this application. (4) Thank you for the suggestions regarding our discussion section. We will adapt our conclusion accordingly, to address limitations of the current method, as well as extend our statement about future work to include clinical next steps containing a retrospective study for optimal cement amount estimation.Comments to Reviewer #1 (7.1) We will give clear definitions of the terms we introduce in this manuscript and adapt Figure 1 to help explicability. (7.2.) We will rewrite the methodology section explaining why we chose specific methods for our framework and how we adapted them. (7.4) We will include videos of the inpainting and spine straightening results in the supplementary material to help 3D visualizations. As for the clinical assessment, this work introduces a retrospective study with plans to extend to a prospective clinical evaluation in future work. As such, clinical results are out of the scope of this manuscript.
Comments to Reviewer #2 (4 & 7.1) We agree with the points raised here. From analyzing our post-operative data we realized that the actual cement injected during surgery was always lower than our upper bound, however there is no correlation between the two as our method was not used for planning. We plan to evaluate this further by extending our dataset and looking at the outcome within a retrospective follow-up clinical evaluation. We will modify the corresponding statements in the abstract accordingly to emphasize volume estimation over amount estimation. (6.2 & 6.3) To avoid confusion related to the vocabulary, we will explicitly explain the meanings of the following terms: “pre-operative”, “post-operative”, “pre-fracture” and “spine atlas”, rephrasing where necessary.
Comments to Reviewer #3 (4.1) The virtual spine straightening algorithm reconstructs the diseased spine’s curvature in all three dimensions to a healthy state, as it maps it to that of a healthy person. In such a way we can indeed correct for kyphosis as a whole. This was perhaps not clearly conveyed and will be rewritten. (4.2) We agree with the reviewer that the likelihood of cement leakage depends also on the kind of fracture. This method is not meant to be used as plug and play, but as a CAI system to support decision making and surgical planning. We will add a comment clearly outlining this.
(7.1) We are currently working on an automatic approach to remove the vertebraes’ spinal processes, thus including only the vertebral body in our analysis. (7.2 & 7.3) Addressing the ethical concerns, we would like to clarify that we carefully went through our dataset retrospectively and identified the rare cases, for which both scans, pre- and post-fracture, exist. All images were taken due to clinical necessity and therefore at this point the sample size cannot be increased.