List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Sutuke Yibulayimu, Yanzhen Liu, Yudi Sang, Gang Zhu, Yu Wang, Jixuan Liu, Chao Shi, Chunpeng Zhao, Xinbao Wu
Abstract
As one of the most challenging orthopedic injuries, pelvic fractures typically involve iliac and sacral fractures as well as joint dislocations. Structural repair is the most crucial phase in pelvic fracture surgery. Due to the absence of data for the intact pelvis before fracture, reduction planning heavily relies on surgeon’s experience. We present a two-stage method for automatic reduction planning to restore the healthy morphology for complex pelvic trauma. First, multiple bone fragments are registered to morphable templates using a novel SSM-based symmetrical complementary (SSC) registration. Then the optimal target reduction pose of dislocated bone is computed using a novel articular surface (AS) detection and matching method. A leave-one-out experiment was conducted on 240 simulated samples with six types of pelvic fractures on a pelvic atlas with 40 members. In addition, our method was tested in four typical clinical cases corresponding to different categories. The proposed method outperformed traditional SSM, mean shape reference, and contralateral mirroring methods in the simulation experiment, achieving a root-mean-square error of 3.4 ± 1.6 mm, with statistically significant improvement. In the clinical feasibility experiment, the results on various fracture types satisfied clinical requirements on distance measurements and were considered acceptable by senior experts. We have demonstrated the benefit of combining morphable models and structural constraints, which simultaneously utilizes cohort statistics and patient-specific features.
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43996-4_31
SharedIt: https://rdcu.be/dnwO5
Link to the code repository
N/A
Link to the dataset(s)
Open Source Atlas of Pelvis Segmentations: https://github.com/I-STAR/PelvisAtlas
Clinical-data-on-pelvic-fractures: https://github.com/Sutuk/Clinical-data-on-pelvic-fractures
Reviews
Review #1
- Please describe the contribution of the paper
The author(s) propose a 2-step method for automatic virtual reduction of pelvic fractures, including complex traumas, such as iliac and sacral fractures, dislocation and combined trauma. The proposed method was tested in a simulation study and in a retrospective clinical study.
- 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 author(s) build on previously published methods for pelvic reduction and proposed a novel two-step method to overcome earlier limitations. Results of the proposed method are compared to previously published method and showed significant improvements.
- 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.
Although the author(s) mentioning that the proposed method is automatic, it does require manual involvement in the bone fragment segmentation.
- 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 manuscript gives sufficient details to reproduce the proposed method. Data for the generation of the statistical shape models are publicly available, and the author(s) also published the data used for the clinical study. I, therefore, believe the work has a very high reproducibility.
- 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
Thank you for the interesting work and the well structured and written manuscript. Please see below a few comments, remarks I have:
- In equation (1): Without reading the referenced work for GPMM, some of the symbols used in the equation and text above are not defined. Including brief explanation for ‘mu’ and ‘k’ will improve the readability of the manuscript.
- The evaluation of the clinical cases seems a bit vague, especially compared to the evaluation of the simulation study. Various previous studies have used surgeon-planned reduction as gold standard to evaluate the image-guided planning. Maybe there was an option to compare the plans with the clinical achieved reduction?
- On a side note (just curiosity on my side): It is well known that the shape of the pelvis has significant differences between male and females. I noticed that the 40 datasets for the atlas generation are equally split between female and male – ensuring that both mean shapes are included. However, I was wondering if anybody every tried to make separate shape models for the female and male pelvis and if this could improve the registration accuracy?
- 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?
To my knowledge the method proposed is novel and build-up on the strength and weaknesses of previously published methods. The method is well tested and achieved significant better results then previous methods.
- 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 #2
- Please describe the contribution of the paper
The paper describes multistep pelvic fracture reduction planning including statistical shape model based symmetrical complementary (SSC) registration to determine the pose of each fragment relative to the healthy pelvic bone structure, followed by detection and matching of sacroiliac joint and pubic symphysis based on Gaussian process morphable models.
- 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 novelty could lie in the use of localizing kernel to detect sacroiliac joint and pubic symphysis and use pelvic geometric symmetry to constraint the structure and joint alignment.
- 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 approach assumes mirror symmetry of a pelvis which may not always be true. Notations are confusing and not well defined. Likewise, the method section is not well organized nor written. It is unclear how optimization of rigid transformation of each fragment and GPMM parameters is done in symmetrical complementary registration. Which is the selected reference? What is the source and what is the target? Does Eq. 4 return a vector or scalar value? Is an approach described in 2.2 nonrigid or rigid registration? For articular surface detection, what is the average normal direction and how was it determined? What are surface landmarks? For the articular surface matching, why a ray-tracing collision detection is needed and what does it yield? What are complementary surfaces? How TD is estimated when it does not appear in any equations?
- 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
Reproducibility of the work is difficult if not impossible since the method is not clearly described.
- 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 paper describes multistep pelvic fracture reduction planning including statistical shape model based symmetrical complementary (SSC) registration to determine the pose of each fragment, followed by detection and matching of sacroiliac joint and pubic symphysis based on Gaussian process morphable models. The novelty could lie in the use of localizing kernel to detect sacroiliac joint and pubic symphysis and use pelvic geometric symmetry to constraint the structure and joint alignment. The approach, however, assumes mirror symmetry of pelvic structure which may not always be true. Notations are confusing and not well defined. Additionally, the method section is poorly organized and written. It is unclear how optimization of rigid transformation of each fragment and GPMMs parameters is done in symmetrical complementary registration. Which is the selected reference? What is the source is what is the target? Does Eq. 4 return a vector or scalar value? Is an approach described in 2.2 nonrigid or rigid registration? For articular surface detection, what is the average normal direction and how was it determined? What are surface landmarks? For the articular surface matching, why a ray-tracing collision detection is needed and what does it yield? What are complementary surfaces? How TD is estimated when it does not appear in any equations? What are feasibility constraints?
- 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?
As mentioned in (6), the paper is not well written, making it hard to reproduce and determine whether the method works.
- Reviewer confidence
Very confident
- [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed
6
- [Post rebuttal] Please justify your decision
The authors addressed questions raised by reviewers well.
Review #3
- Please describe the contribution of the paper
The authors propose a two-stage method to automatically reposition bone fragments and joint dislocations of pelvic structures using a shape model based registration method. The authors tested the method on simulated pelvic fractures, compared their approach to different SOTA methods as well as examined four clinical cases.
- 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.
I believe it is a well-structured paper with an interesting approach for a complicated problem in trauma surgery. It is nice to see a strong paper from the non-DL domain.
- 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.
I regard it as a weakness that the segmentation problem here is assumed to be solved. To fully utilize the approach in the clinical setting, the approach would need to be robust towards segmentation errors. I appreciate that the authors considered this as a regularization in their optimization, but it is not evaluated. Of course with the recent successes in DL based segmentations, we can now create high quality segmentations, but I believe the fragments of pelvic fractures may lead to inaccuracies. So I think the method and the claims would be stronger by completing the whole workflow and evaluating it. Furthermore, 4 retrospective clinical cases are a bit limited, although I understand that evaluation takes some time. The comparatively high time consumption (excluding the manual or automatic segmentation) is not discussed as part of the clinical workflow. There are no limitations mentioned.
- 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 data will be published but not the code. It is difficult to say how easy it will be to reproduce the exact results by re-implementing the GPMM. Statistical test were provided along with boxplots showing the error distributions.
- 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
- Have you considered mirco-fragements of the bones? Would this be a problem for the model?
- Have you experimented with inaccuracies of the segmentation?
- Which part of the modelling does take the most time of the method? Could that be improved?
- Can you describe the difference / domain gap between the simulated and clinical data?
- Would it be possible to include the segmentation step in the modeling in a shape&appearance model to overcome the segmentation step?
- Please discuss the failure cases in more detail and possible limitations
- How would the method be incooperated into the clinical workflow?
- 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 grade is given based on the mentioned weaknesses.
- 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
6
- [Post rebuttal] Please justify your decision
My major comment on the impact of segmentation errors has been addressed. The reproducibility still could be a problem without releasing any codebase, so I would advise to do so just like it was done with pelvis dataset.
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 manuscript received mixed reviews, requiring clarifications.
Strenghts of the work include an interesting method to an important problem that outperforms previous approaches. Some components of the method are perceived as novel.
The concerns include:
- Some reviewers had considerable difficulty extracting crucial information from the manuscript, while others perceived the manuscript as well structured and presented. Some clarifications here seem necessary.
- The other echoed weakness includes claims around the method being automated. There is, however, an assumption that pelvis segmentation can be performed in an automated fashion, which is not considered here. Claims should be adjusted.
- Further, there is concern that the symmetry assumption may not hold in practice.
Author Feedback
We thank the reviewers for their constructive comments. We have modified the manuscript accordingly and would like to make the following clarifications.
In our method, pelvic reduction planning is solved in two stages. In the first stage, GPMMs are used to model ilium and sacrum, and fractures within each bone are addressed by SSC registration. In the second stage, ilium and sacrum are aligned by the AS matching method to restore joint dislocations and thus the whole pelvic morphology.
GPMM: We use GPMM to model the point cloud of pelvic bones. GPMM models deformation as a Gaussian process with a mean function µ and covariance function k, and is invariant to the choice of reference. Therefore, the reference can be chosen arbitrarily.
SSC registration: A bilateral supplementation strategy is used in model adaptation: mirrored counterpart are algined and merged to the target point cloud to provide additional guidance. Specifically, fragment alignment and model adaptation are performed alternatingly: the fragment (source1) and the mirrored counterpart (source2) are first rigidly aligned to the GPMM model (target); then the GPMM is non-rigidly deformed towards the merged point clouds (source1+2). The output deformation vector field from Eq. 4 is used to drive the model adaptation in Eq. 5. The iterations continue until convergence.
AS detection and matching: As indicated by the red and blue regions in Fig. 1 - AS detection, surface points in the joint regions are annotated as landmarks. They are first annotated in the model template and then propagated to each instance using a GPMM model (with non-rigid adaptation). Each landmark (vertex in mesh data) is associated with a normal vector, and the average normal vector of joint landmarks is computed as the “average normal direction”, which is then used to eliminate outliers. Complementary surfaces refer to pairs of corresponding joint surfaces (e.g., the left and right sides of the pubic symphysis). In AS matching, a collision-detection-based term is employed as a feasibility constraint in Eq. 6 to ensure that surfaces do not overlap. The collision detection returns the set of points collided, which is then used to compute the IoU. The cost function in Eq. 8 is optimized with respect to Td, which determines the pose of the moving bone and its joint surfaces. We appreciate the reviewers’ valuable advice and have incorporated the clarifications into the manuscript to improve readability.
Fracture segmentation: A fully automatic planning pipeline does require a fracture segmentation module. We have developed a DL pelvic fracture segmentation method in a separate study, which has already been accepted for publication. We hope the synergy of the two works could bring greater impact and are eager to discuss the influence of segmentation results and the overall workflow in future study.
Symmetry assumption: Although pelvic structure is not always symmetric, a recent study (Krishna et al, 2022) has pointed out that the symmetry-based method outperformed SSM. Therefore, our method aims to combine contralateral information with statistical model to benefit from both, and does not assume strong symmetry. The validity of this rationale is proved by the improved planning accuracy in the experiment.
Clinical evaluation: Due to the limited space, we did not perform quantitative comparisons against other methods on clinical data. We will combine geometric measurements and manual planning for a comprehensive evaluation upon a journal version.
Limitations: The CMA-ES optimization is the most time-consuming step in the workflow. We have accelerated this process by CPU parallelization. We plan to improve the loss function design and computation, including a more efficient collision detection method. We also observed suboptimal results on cases where both sacroiliac joints are dislocated, which are relatively rare in clinical practice. We plan to focus on addressing this type of injury.
Post-rebuttal Meta-Reviews
Meta-review # 1 (Primary)
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
All reviewers are in agreement that the manuscript can now be accepted. However, the reviewers emphasize the need for clarification of several aspects in the manuscript which need to be taken into account for preparation of the final version.
Meta-review #2
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
The paper presents a novel two-step method for automatic virtual reduction of pelvic fractures, leveraging Gaussian Process Morphable Models (GPMM) and a unique symmetrical complementary registration approach. In reconciling the review comments, it was clear that all reviewers found the contribution significant but had concerns around the clarity of explanation, the reliance on manual segmentation, and the assumption of symmetry. The authors’ rebuttal clarifies many points and notes improvements in the revised manuscript. The authors also reference another accepted paper addressing segmentation, hinting at a full, automated solution in the future. The key concern that remained unaddressed is the relatively limited evaluation on retrospective clinical cases and lack of a comparison to the clinical achieved reduction. This is a valid point to consider and should be addressed more thoroughly by the authors. In conclusion, the paper is promising and could have a significant impact on trauma surgery. However, further efforts on evaluation, improved clarity in the method section, and a more comprehensive automated workflow would significantly enhance the paper’s contribution.
Meta-review #3
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
After rebuttal all the reviewers are in agreement that this paper is worthy of publication in MICCAI