Abstract

Accurate and reliable selection of the appropriate acetabular cup size is crucial for restoring joint biomechanics in total hip arthroplasty (THA). This paper proposes a novel framework integrating square-root velocity function (SRVF)-based elastic shape registration technique with an embedded deformation (ED) graph approach to reconstruct the 3D articular surface of the acetabulum by fusing multiple views of 2D pre-operative pelvic X-ray images and a hemispherical surface model. The SRVF-based elastic registration establishes 2D-3D correspondences between the parametric hemispherical model and X-ray images, and the ED framework incorporates the SRVF-derived correspondences as constraints to optimize the 3D acetabular surface reconstruction using nonlinear least-squares optimization. Validations using both simulation and real patient datasets are performed to demonstrate the robustness and the potential clinical value of the proposed algorithm. The reconstruction result can assist surgeons in selecting the correct acetabular cup on the first attempt in THA, minimising the need for revision surgery. Code and data are available at: https://github.com/zsustc/3D-ASR

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/1272_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/zsustc/3D-ASR

Link to the Dataset(s)

https://github.com/zsustc/3D-ASR

BibTex

@InProceedings{ZhaShu_3D_MICCAI2025,
        author = { Zhang, Shuai and Wang, Jinliang and Wang, Xu and Konan, Sujith and Stoyanov, Danail and Mazomenos, Evangelos B.},
        title = { { 3D Acetabular Surface Reconstruction from 2D Pre-operative X-ray Images using SRVF Elastic Registration and Deformation Graph } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15975},
        month = {September},
        page = {2 -- 11}
}


Reviews

Review #1

  • Please describe the contribution of the paper

    The paper’s main contribution is applying [18] and integrating square root velocity transformation into the framework for the targeted clinical application - estimating the correct cup size for THA by registering the extracted contours from fluoroscopy images to the 3D implant model. The goal is to reduce radiation usage using three fluoroscopy images instead of a 3D scan.

  • Please list the major strengths of the paper: you should highlight 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.
    • A comprehensive introduction and motivation
    • Robustness testing and clinical data verification
    • An innovative approach to using square root velocity transformation for deformation graph optimization
  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
    • The mathematical derivation is non-trivial to parse. Square-root velocity transformation is not introduced before use.
    • It is unclear how the last term E_obs (Eq 5) is related to Eq 3. Eq 3 is defined in 2D (O \in SO(2)), but the Eq 5 is in 3D (R_k \in SO(3)). \phi is also not defined.
    • Inconsistent results. In Table 1, the measurement of MAEs is from 0.98mm to 1.63mm, but the text says 0.98mm to 1.3mm. Similarly, the measurement of SDs is 0.95mm to 1.61mm in the table, but it says 0.95m to 1.25mm in the text.
    • Missing baseline comparison. The novelty of this approach is the use of the square root velocity transformation. The authors should compare the results with and without using the transformation, i.e., what happens if [18] is used directly.
  • Please rate the clarity and organization of this paper

    Poor

  • Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.

    The authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html

    N/A

  • Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.

    (3) Weak Reject — could be rejected, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    This paper extends [18] by incorporating square root velocity transformation to optimize the deformation graph for fitting the extracted contours from fluoroscopic images to the 3D implant model. The approach is innovative, but the paper could be improved in the following areas: 1) Introduce the square-root velocity transformation before using it. 2) Make the mathematics description more rigorous. For example, In Section 2, the second paragraph, in the second line, N_P is not defined. Is it the number of implant model points? In \tilde{C}k, the last point is \tilde{P}{k, N_k^{M}}. What does M mean here? In Section 3, N_E is the number of graph nodes, but it never mentions which N_E is used in the experiment. Is N_E the same as N_P? If not, how is it “uniformly” downsampled? Is the graph a grid? Mathematics does not describe the graph’s connectivity, but it directly uses “m-neighboring nodes.” Usually, the node’s neighborhood is defined by the graph connectivity. Yet, here, it seems the “neighbors” are defined using distance between nodes. If there is no “connectivity”, it is not actually a graph. It is more like a particle system. 3) The discrepancy between Table 1 and the text under it should be clarified. 4) In Section 4, there is a justification for which methods were excluded from the experiments. While the justification is understandable, comparing the proposed method with the state-of-the-art is hard. Are there existing SSMs/ML models available for the application? 5) Given the main contribution is the new term E_obs in Eq 5, it would be nice to see its effect. The authors have not mentioned what w_rot, w_reg, and w_obs were used in the experiment. For a new proposed energy, it would be desirable to see an analysis of its weighting parameters. 6) If possible, the effect of the number of X-ray images should be studied and reported too.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    Accept

  • [Post rebuttal] Please justify your final decision from above.

    The authors explained the manuscript well and addressed my concerns and questions. Besides the said revision about the mathematics, I encourage the authors to include some of the rebuttal’s clarifications/explanations or make them more explicit in the manuscript so that readers can better interpret and understand the rationale behind the method and evaluation. For example, which experimental parameters were used, why we need SRVF in ED, and how the connectivity of the graph/grid is defined, etc. One question I still have is the definition of the neighborhood. Is it based on connectivity or distance? However, readers can find the answers once the source code is published. Yet, it would be nicer if the manuscript is self-contained.



Review #2

  • Please describe the contribution of the paper

    This paper proposes a method to reconstruct the acetabular surface from 2D X-ray images, to determine the optimal size of the cup for Total Hip Arthroplasty. A hemispherical is projected on the 2D views, then non-rigidly and iteratively deformed to fit the manually delineated acetabulum contour. Specific methods are used to fit the countours (SRVF), in particular. An experimental study proved the accuracy and robustness of the method, then a study on 5 patients shows its application with real data with good results.

  • Please list the major strengths of the paper: you should highlight 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 method is sound and overall well described. I have especially appreciated the SRVF method to register the 2D curves, proven robust by your results, and the ED reinitialization to prevent error accumulating.
    • both experimental studies are well conducted
    • relevant clinical problem to address
  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
    • method: some method points are missing, especially how X-ray poses are estimated in the OR (especially if a CT is not available)
    • but my main concern is the rationale of the approach itself. The surface of the acetabulum is first reconstructed very accurately using a complex method, before you fit an hemispherical model to find the cup size. Is the accurate reconstruction really needed? Why don’t you directly and rigidly fit a non-deformed cup model to the 2D contours? In particular, cup sizes vary in increments (typically of 2mm, but this varies from company to company), so the size value could be one of the parameters to optimize (if not the only one).
  • Please rate the clarity and organization of this paper

    Satisfactory

  • Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.

    The authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
    • as stated, cup size does contribute to THA complications; but to what extent? Although it is reported that there are more dislocations with larger cups, and that experienced surgeons may tend to use smaller ones, I don’t know how important cup size is among many factors (length of neck, angle, …).
    • that being said, the motivation for cup size estimation from 2D X-ray images only is well founded

    • Method
    • “… where the poses of N views of X-ray images assumed to be known.” -> This is a very strong assumption. In clinics, there will be considerable uncertainty. If I understand correctly, 2D-3D registration with the segmented CT model is used in your patient study (p.8) to estimate these poses. If so, this step should be included as part of the method. Is CT segmentation required? What would you propose if no CT is available?
    • “Based on the reconstructed acetabular surface models, the acetabular cup size is estimated by fitting hemispherical surface models, with the diameter of the fitted models used as the cup size.” -> How is this fit computed? This should be part of the method, not at the end of the results section. Also it could be more detailed.

    • Simulated data
    • good study, yielding to satisfactory results even with increased noises

    • Patient data
    • in your experiments, did you calibrate the X-ray scanner? Or do you assume theoretical, ideal perspective parameters?
    • contours were segmented by two surgeons. Did you run the registration twice to estimate differences, or did you agregate both contours (average, by selecting one, or by modifying one according to both observations, …)? This should be detailed.
    • p.8 “The poses of X-ray images used in our reconstruction validation are obtained from 2D-3D registration to CT-segmented pelvic models.” -> In a real scenario, will this registration also need to be performed? How accurate is the pose estimation? Even though your method has proven to be quite robust, pose inaccuracy can be a major challenge.
    • “Compared to the intra-operatively confirmed cup size, the estimation error is 1.3 mm, with a standard deviation of 1.26 mm.” -> This could have been developed since determining cup size was your primary goal. Given the existing cup sizes, would your method have led to the selection of the final size eventually used?
    • detail in Fig.3: for the 4th case, 2nd X-ray image, the observation is missing which mix the color code. If possible, input and reconstruction should remain green and red for consistency.
  • Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.

    (4) Weak Accept — could be accepted, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    Interesting method and good results regarding the acetabular surface reconstruction. However, determining the size cup itself is very indirectly addressed, which may not be optimal.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    Accept

  • [Post rebuttal] Please justify your final decision from above.

    The rebuttal letter is strong and addresses most points of concern, both methodological and regarding the application itself. Providing the camera-ready paper integrates the proposed clarifications, this paper -if accepted- will be a valuable contribution to the MICCAI community.



Review #3

  • Please describe the contribution of the paper

    The paper proposed an approach for Acetabular Surface Reconstruction for implant size estimation before total hip replacement surgery. From manual segmentations of three pre-operative 2D X-ray images, the 3D acetabular surface shape can be accurately reconstructed. The methodology is based on square-root velocity function and embedded deformation.

  • Please list the major strengths of the paper: you should highlight 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 a novel application of approaches from computer graphics (square-root velocity function and embedded deformation) to the medical field. The approach is clinically relevant, it can save time, radiation dose, and help reduce the number of revision surgeries due to wrong choice of acetabular cup size.

    The authors provide relatively clear descriptions of their methodology and evaluation. Opposed to previous approaches the method is compared to, the proposed approach is robust to various sources and levels of noise in the data.

  • Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.

    The method description is very detailed and the processing pipeline is nicely visualized in Fig. 1. However, the applied square-root velocity function method is never explained, and no reference is given for the reader to understand how this crucial part of the proposed method works (from a quick literature search I think [1, 2] should be cited). Adding a short description would make it easier to understand the proposed method.

    A very detailed presentation of the results is given, but a more critical discussion would strengthen the paper. Fig. 2 rightmost column should be updated such that bars aren’t cropped or only in a meaningful way. Fig. 3 middle column 4th row image needs to be updated. While the metrics show an improvement and give error ranges both on the surface reconstruction and the cup size estimation, some more context would be helpful to understand if these results are precise enough. If the implant cup sizes are available in 5 mm increments, a 1.3mm +- 1.3 mm is great, if the increment is in the sub-millimenter range, the estimation would need improvement to be clinically meaningful.

    It would be very helpful to see a direct comparison between the state of the art statistical shape model-based approaches and the presented work. I understand the explanation the authors give as to why they were not able to perform this comparison. Still, in the discussion of their results the authors should comment on how the results compare to this method (error magnitudes, speed, manual interaction).

    [1] S.H. Joshi, E. Klassen, A. Srivastava and I.H. Jermyn, “A Novel Representation for Riemannian Analysis of Elastic Curves”, Proc. IEEE Conf. Computer Vision and Pattern Recognition, pp. 1-7, 2007.

    [2] A. Srivastava, E. Klassen, S. H. Joshi and I. H. Jermyn, “Shape Analysis of Elastic Curves in Euclidean Spaces,” in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 7, pp. 1415-1428, July 2011.

  • Please rate the clarity and organization of this paper

    Satisfactory

  • Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.

    The authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html

    I haven’t worked with square-root velocity function and embedded deformation before.

  • Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.

    (4) Weak Accept — could be accepted, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    This contribution can be clinically relevant, and the presented descriptions and evaluations are clear and sound. The paper is missing an explanation of the square-root velocity function method, a main part of the proposed method, and a more critical discussion and interpretation of the results would strengthen the paper.

  • Reviewer confidence

    Somewhat confident (2)

  • [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.

    Accept

  • [Post rebuttal] Please justify your final decision from above.

    The authors addressed my three major concerns in their rebuttal. If they add the promised explanations of their method, update the figures, and give the reader more understanding of the impact of their results, they will have responded to two of my three main concerns.

    I still think a comparison to state of the art statistical shape model-based approaches would strengthen the paper decidedly. Even if the authors can not perform a direct comparison, giving a context/appraisal (e.g., error measures of SSMs on healthy bones) would be very helpful for the reader to understand how good the presented results are.

    If the other reviewers are content with the rebuttal and agree, I think this contribution can be accepted.




Author Feedback

We thank reviewers for their constructive feedback. We are encouraged that our method is regarded as “novel” (R2, R3), “sound, well, appreciated, robust, clearly described, and address a clinical problem” (R1, R3), the “experimental studies are well conducted and obtain satisfactory results” (R1), and that there is “robustness testing, and clinical data verification” (R2). We address below the main criticism.

<R2, R3: Comparison with SSMs> SSMs reconstruct “statistically normal” anatomy thus are unsuitable for cases with partial or extensive acetabular bone loss. In THA, key anatomical landmarks (rim and center) are missing/distorted. SSMs would inherently smooth out pathological defects rather than preserve them. Therefore, SSM-based fitting is unreliable for our patient-specific reconstruction.

<R2, R3: SRVF is not explained, R2: math notations need clarification> We will describe SRVF earlier, clarify notations and description, add suggested reference.

<R2: Work is extension of ED method [18], and direct comparison> We respectively disagree as ED isn’t fully automated. It is a general-purpose mesh deformation tool, requiring explicit positional constraints which our SRVF registration provides. Without SRVF, ED alone is under-constrained and fails, so a “no-SRVF” ablation isn’t a valid baseline.

<R2: Method description non-trivial> While R1 and R3 noted that the method is sound, well-described and clear, we appreciate R2’s comment and clarify that: Following [18], we construct the ED graph as a valid graph, the mesh is voxelized using a fixed grid size, the centroid of each voxel’s vertices is used to uniformly sample ED nodes. Each node is explicitly connected to its neighbours by undirected edges. These represent their shared regions of influence and introduce the regularization term in Eq.4, enforcing consistency between the affine transformations of connected nodes. Function \phi (Eq.1) projects 3D vertices of the deformed model onto the 2D image plane using known poses (R_k, t_k). Eq.3 establishes correspondences between observed and model‐projected contours. Eq.5 uses these to link each 2D observation back to its matched 3D vertex, projects that vertex via \phi, and penalizes the 2D reprojection error. The weights in Eq.4 are set with w_{rot} =1, w_{reg}=10, w_{obs}=100. We tested 2–6 simulated images with consistent noise and found that 2 views, separated by a 20° angle, reduced accuracy. Using 3 views yielded performance similar to 4-6 views.

<R1: X-ray pose accuracy, acquisition in the OR> X-ray poses were obtained by rigidly registering CT-segmented pelvic models to 2D X-ray contours. Our framework itself does not require CT data. In practice, X-ray poses can be obtained through built-in encoders or sensors (in modern C-arms) and optical tracking (IR markers on the C-arm and patient). Approximate intrinsics were read from the x-ray header files. Contours were obtained by two surgeons and reviewed to reach consensus. We have experimentally validated that our approach remains robust under moderate pose perturbations, ensuring reliable results in practice.

<R1: Fit a non-deformed cup model to the 2D contours> Directly fitting a non-deformed cup model to 2D contours overlooks patient-specific defects (rim erosion, bone loss, wear) which cause significant deviations from a perfect sphere. Our 3D reconstruction captures these irregularities, enabling not only more accurate cup sizing, but also critical tasks like defect quantification, graft planning and optimal implant orientation. Crucially, our method allows to first assess the reconstructed defect for guiding filling, and then simulate joint motion (e.g., impingement) to determine the optimal implant size. We will highlight this.

<R1, R3: cup size selection> With standard implants in 4 mm increments, our MAE of 1.3mm (SD 1.26) stays within ±2 mm half-step and the predicted cup size matched the final size selected in surgery in all tested cases.




Meta-Review

Meta-review #1

  • Your recommendation

    Invite for Rebuttal

  • If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.

    N/A

  • After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.

    Accept

  • Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    N/A



Meta-review #2

  • After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.

    Accept

  • Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    All reviewers agree for acceptance of the work after rebuttal. The authors have also done a good job in the rebuttal.



Meta-review #3

  • After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.

    Accept

  • Please justify your recommendation. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    After thye rebuttal, all three reviewers recommend acceptac of the paper as the authors have addressed the three major concerns.



back to top