List of Papers By topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Morgan Ringel, Jon Heiselman, Winona Richey, Ingrid Meszoely, Michael Miga
Abstract
Image-guided surgery requires fast and accurate registration to align preoperative imaging and surgical spaces. The breast undergoes large nonrigid deformations during surgery, compromising the use of imaging data for intraoperative tumor localization. Rigid registration fails to account for nonrigid soft tissue deformations, and biomechanical modeling approaches like finite element simulations can be cumbersome in implementation and computation. We introduce regularized Kelvinlet functions, which are closed-form smoothed solutions to the partial differential equations for linear elasticity, to model breast deformations. We derive and present analytical equations to represent nonrigid point-based translation (“grab”) and rotation (“twist”) deformations embedded within an infinite elastic domain. Computing a displacement field using this method does not require mesh discretization or large matrix assembly and inversion conventionally associated with finite element or mesh-free methods. We solve for the optimal superposition of regularized Kelvinlet functions that achieves registration of the medical image to simulated intraoperative geometric point data of the breast. We present registration performance results using a dataset of supine MR breast imaging from healthy volunteers mimicking surgical deformations with 237 individual targets from 11 breasts. We include analysis on the method’s sensitivity to regularized Kelvinlet function hyperparameters. To demonstrate application, we perform registration on a breast cancer patient case with a segmented tumor and compare performance to other image-to-physical and image-to-image registration methods. We show comparable accuracy to a previously proposed image-to-physical registration method with improved computation time, making regularized Kelvinlet functions an attractive approach for image-to-physical registration problems.
Link to paper
DOI: https://doi.org/10.1007/978-3-031-43996-4_33
SharedIt: https://rdcu.be/dnwPc
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 paper describes a model that formalizes linear elasticity and proposes a regularization method to use it for image-to-physical registration.
- 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 addresses a very interesting topic of image-to-patient (or image-to-physical) registration that is essential to enable IGS and that can be employed in many surgical applications
- The paper proposes the formulation of a nonrigid point-based deformation using what is called “grab” and “twist” deformations in a linear elastic model
- The experiment is well designed
- The method has been carachterized/calibrated using volunteers data and validated on real surgical data of a pathological patient. Although only one use case was reported, this is a good step towards clinical acceptance
- The achieved accuracy is similar to state of the art FEM method
- Computation time is 14s which demonstrates feasibility in surgery
- 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.
Maybe one weakness of the paper is the lack of discussion regarding clinical relevance of the results for the proposed method vs. FEM or even image-to-image registration, especially regarding accuracy, whether this is sufficient for IGS navigation for breast cancer surgery, and what is/should be accepted/expected in the operating room. Also the authors should include a small discussion regarding the role, accuracy, and timing for acquiring fiducial landmarks during surgery and what is the impact of inaccurate landmarks on the registration outcome
- 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
Good
- 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
Minor comments:
- Abstract needs to include a summary of quantitative results
- In section 4, the text states that the registration is “real-time”, which I would argue that it is not true. I believe 14s of computation time is widely accepted in a surgical workflow, but won’t work for a real-time application
- 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 topic addressed is relevant to MICCAI community
- the deformation model proposed is an original way of performing nonrigid registration and could be interesting for similar applications
- the paper is well organized, the experiment is thoughtfully designed and validation carried out on clinical data
- 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
7
- [Post rebuttal] Please justify your decision
- The authors provided sufficient justifications for the significance/relevance of the achieved accuracy in a clinical application
- Although, the impact of fiducial errors was acknowledged, the authors response does not include any additional information that address this concern in the rebuttal.
- I still find the use of “real-time” misleading, as this is usually employed for specific applications that require continues registration, for example.
Review #2
- Please describe the contribution of the paper
This work address an important topic in computer assisted intervention which is the modelling of breast deformation. It proposed a closed-form solution based on Kelvinlet functions by assuming the linear elasticity of the breast tissue. By doing so, the problem space is simplified, consequently reduced computational cost. Quantitative evaluation are conducted to compare the proposed method with other state-of-the-art 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.
This paper is well written and very clear structured. Especially the nomenclature and math expression are used consistent and derived/described clearly. In terms of novelty, this work applied an state-of-the-art method which is commonly used in computer graphics to simulate the deformation of rubber-like objects, adapting it to medical field. For homogenous elastic materials, Kelvinlet functions is a very efficient tool to model the deformations. In the evaluation, real-world data are used. In particular, both deformation generated under instruction as well as deformation resulting from real clinical workflow are used. It can be concluded that the set up of the experiments are carefully designed and well suited for this problem statement. The result shows that the proposed method can be considered as an alternative to other state-of-the-art methods
- 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 main weakness of this paper come along with the basic assumption of Kelvinlet functions. It is originally proposed to simulate the deformation of rubber-like linear elastic materials. Although it is very simple and has a closed-form solution, its adaptation in the medical field is limited, because tissues are rarely homogenous rubber-like materials, organs with tumors ore calcifications are known to have heterogenous biomechanical properties. In other words, it lack of accuracy while simulating the deformation of complex organs/tissue, and probably also not suitable for treatment planning and guidance with high requirement on the precision. The authors stated these limitations in the last section clearly.
- 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 is moderate. source code and data set are not provided. However, the method and experiment set up are described very clear, therefore the reproduction of the algorithm and the experiment are possible.
- 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
A minor point which I am missing in this paper is the justification of the selected mechanical breast properties (v, E, w_E). Are there any clinical evidence justifying this choice? Or any reference work?
- 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?
Well structured and well written. Novelties and limitation are clearly demonstrated
- 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 method for nonrigid registration of the human breast, which involves image-to-physical registration using “grab” and “twist” regularized Kelvinlet functions. The sensitivity of the regularized Kelvinlet function hyperparameters is explored on a supine MR breast imaging dataset obtained from 11 patients. Furthermore, the approach is validated on a dataset from a breast cancer patient. Although the registration accuracy of this method is not as good as that of state-of-the-art methods, it is capable of achieving near real-time performance.
- 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.
1.The clinical significance of the study has been well illustrated. 2.The manuscript is well-written and easy to follow. 3.Based on the experimental results, the proposed method outperforms previous methods with regards to registration time, while achieving comparable accuracy.
- 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 proposed method’s novelty may not be significant enough. In my perspective, the most critical element of the proposed framework is based on the regularized Kelvinlet functions for volumetric digital sculpting, as mentioned in the work referenced in [14]. 2.The validation and discussion of the proposed registration method may not be sufficient to prove the clinical value of the method.
- 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 reproducibility depends on whether the code and dataset are to be open sourced.
- 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
Firstly, I would like to express that the paper is well-written. However, I have some doubts regarding the clinical applicability of the proposed method. In my opinion, the potential advantage of using this method in clinical practice was not discussed sufficiently, especially given that the achieved registration accuracy is not satisfactory. Thus, further validation and improvement are needed to establish its practical clinical value. Additionally, it is important to emphasize that the use of skin fiducials is one of the limitations of this method. Furthermore, the legends in Figure 3 are unclear and could benefit from clarification. Lastly, in the second paragraph of the Introduction section, it would be helpful to either define “FEM” before using the abbreviation or replace it with its full name for clarity。
- 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?
The paper is well-structured and well-written overall, and the proposed method is explained clearly. The method is capable of achieving near real-time performance, which could be beneficial for improving navigation during image-guided surgeries if its accuracy can be further improved.
- 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
4
- [Post rebuttal] Please justify your decision
Thanks for the authors’ rebuttal. However, from my perspective, the authors’ rebuttal comments have failed to address my concerns regarding the methodological issues and clinical applications identified in this study.
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 submission tackles the problem of modelling breast deformation with a linear elastic model using regularized Kelvinlet functions. The main motivation is to improve registration methods that compensate for soft-body deformation, with respect to both speed and accuracy.
This submission received mixed reviews: R1 - strong accept, R2 - accept, R3 - weak-reject. Some important weaknesses have been raised, and for this reason, I recommend that the authors are invited to respond to these criticisms in a rebuttal.
The authors should carefully consider all the reviewer comments, and address the main negative comments in the rebuttal, which can be broadly categorised as follows:
- Lack of technical novelty (R3 - item 6). In your response, state clearly why this work is not a simple application / basic adaptation of [14] to a different anatomical target (breast tissue).
- Inadequate validation for demonstrating clinical value (R3 - item 6), and inadequate accuracy for clinical application (R3 - item 9)
- Practical limitations concerning the use of fiducials (R3) and lack of discussion about these issues in the clinical setting, and risks (R1). Additionally, no discussion about risks / method impact associated to imprecise landmark placement (R1)
- Lack of discussion / motivation concerning the clinical relevance of the method compared to existing works, especially concerning target accuracy (R1)
- Lack of justification and clinical validity for the use of Kelvinlit functions - related to fundamental assumptions concerning heterogeneous biomechanical properties (R2 - item 6)
- Lack of justification for mechanical parameter selection (R2 - item 9)
- Unjustified ‘real-time’ claim (R1)
Author Feedback
We thank the reviewers for their comments. The critiques listed by the meta-reviewer are addressed below.
Lack of novelty: Although regularized Kelvinlets (RK) are derived and presented in [14], we propose RK as a deformation basis for a medical imaging registration task. [14] presented RK for computer animations in a forward-solve manner. This work is the first to use RK in an inverse-problem to reconstruct a deformation state based on sparse data inputs. This is a non-obvious extension of [14] and represents technical novelty.
Inadequate validation and accuracy for clinical application: This work presents an experimental framework to measure target registration error (TRE) in breast tissue. The breast imaging dataset was acquired specifically for validation – it allows for ground-truth measures of subsurface displacements, enabling error measures unachievable in an intraoperative setting. For clinical context, current breast guidance technologies localize a single tumor-implanted seed. No intraoperative information about the tumor boundary is provided. As a result, resections can have several centimeters of tissue beyond the cancer margin. Despite seed information and large resections, reoperation rates are still high (~17%) emphasizing the need for additional guidance technologies [1]. With respect to our reported accuracy, breast image guidance does not require sub-millimetric accuracy. What it does require is an understanding of accuracy such that appropriate borders can be communicated to the surgeon to ensure uninvolved margin while also maintaining cosmesis. Furthermore, studies suggest that ~10 mm margins beyond an MR-visible lesion are appropriate for effective radiation therapy, which is the next step of breast conserving therapy [2]. Therefore, the reported TRE (3.0 ± 1.1 mm) is sufficiently accurate to add value considering the clinical context. We will include this motivation and discussion in Sections 1 & 4. [1] Kaczmarski et al, J Am Coll Surg 2019. PMID: 30703538 [2] Schmitz et al, Radiother Oncol 2010. PMID: 20826026
Use of fiducials: While the need for fiducials may be a limitation for some applications, we do not anticipate it being a barrier in breast surgery. Previous work utilized ink-markings (often used in breast surgery) to mark fiducial points after imaging so that fiducials can be removed from the surgical field while maintaining point correspondence between imaging data and the breast surface [3]. Similar fiducial placement and localization strategies in neurosurgery do not significantly extend procedure time. Errors from imprecise fiducial placement were not reported, but skin surface fiducial tracking is explored extensively in [3]. We will include this discussion in Section 4. [3] Richey et al, IEEE Trans Biomed Eng 2023. PMID: 37018246
Heterogeneous properties: We recognize the limitation of the RK homogeneous and isotropic assumptions in Section 4. However, we note that heterogeneous and anisotropic models only improve TRE of breast registration by <1 mm [4]. While heterogeneous models are more biomechanically accurate, they are computationally expensive and require boundary condition inputs that are very challenging to infer intraoperatively. Considering the advantages of near real-time registration speed and compatibility with sparse data inputs, RK are appropriate for modeling intraoperative breast deformations. [4] Ringel et al, Clin Biomech 2023. PMID: 36890069
Mechanical parameter selection: Parameter selection was based off reported literature values [5,6]. We will add citations. [5] Richey et al, IEEE Trans Biomed Eng 2022. PMID: 35604993 [6] Griesenauer et al, Phys Med Biol 2017. PMID: 28520556
Real-time claim: We agree that real-time video rate registration was not demonstrated. However, we use the term “near real-time” to indicate intraoperative feasibility with registration taking 14 s. We do not claim that our algorithm is “real-time” (only “near real-time”) in the text.
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.
This work presents an approach to non-rigidly register human breasts for intraoperative image-guided breast surgery. The method works by placing 26 skin fiducial markers over the breast, then scanning the MR breast ‘preoperatively’ and ‘intra-operatively’. Next, a deformable registration is performed using a mechanical model featuring Regularized Kelvinlet Functions (modelling homogeneous elastic tissue), driven by virtual forces that align the skin fiducials.
The reviews were mixed (R1 - strong accept, R2 - accept, and R3 - weak reject). The main criticisms from R3 were two-fold. First, clinical viability: the application of skin markers for pre-operative MR, which must remain on the skin until the operation, is extremely limiting. I agree with this. Second, the registration accuracy (TRE) is far from clinical needs.
I will also add that the problem being solved here is a classical one studied for a long time in CAI: that of fitting a deformable physics model using sparse point correspondences. In this regard, comparisons with baselines are very insufficient. Only one FEM model is compared (using finite element inverse modeling) from [15], and the proposed technical novelty is to adapt this approach using Regularized Kelvinlet Functions. However, this is a classical registration task: fitting a 3D elastic model using a set of 3D point correspondences. A large variety of elastic registration methods can be applied to solve this task. There is no reason to believe why the proposed Regularized Kelvinlet Functions are better than St. Venant–Kirchhoff (VK) elements, for instance, or various models from the graphics literature such as ARAP [X1]. See e.g. [X2] for a literature review of many models that could be applied to this task,
I therefore find the comparison to a single FEM approach completely inadequate. The main proposed improvement over [15] is the runtime (achieving registration in about 15 seconds). Yet, modern methods exploiting GPUs can perform registration at much faster rates (e.g. using SOFA, designed for real-time interactive surgery simulation), so the relevance of this speedup is in serious doubt. The combination of limited clinical practicality (skin markers), high registration error, and insufficient comparison with various other elastic registration methods used for other organs, means I cannot recommend this for acceptance.
[X1] As-rigid-as-possible surface modeling, Olga Sorkine, Marc Alexa, Proceedings of the fifth Eurographics symposium on Geometry processing [X2] A Review of Deformation Models in Medical Image Registration’ by Monan Wang & Pengcheng Li.
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.
I would lean towards rejecting this paper as some of the valid concerns brought up by the reviewers are still not addressed in the authors rebuttal.
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.
From my perspective, the rebuttal addresses well the most important concerns identified by the primary AC during review, and highlights changes that will be made in revision. Overall, while there is still disagreement between reviewers, given the information available to me the concerns of R3 have been addressed adequately within the scope of a rebuttal period.