Paper Info Reviews Author Feedback Meta-Review Back to top
List of Papers Browse by Subject Areas Author List

Abstract

Statistical shape analysis (SSA) is a powerful tool for studying anatomical structures and their geometric variations in medical imaging. In this work, we analyze real MRI-derived data to explore correlations between geometric deformations and Joubert syndrome (JS). Building on prior SSA research, we tailor the preprocessing pipeline to an in-house dataset and perform a detailed shape variability analysis using principal component analysis (PCA). A random forest classifier is then applied, achieving high classification accuracy. To ensure robustness, we test multiple train-test splits and evaluate their impact. In addition, we support clinical interpretation by providing visualizations that combine 3D and 2D information, resembling typical diagnostic paradigms on MRI planes. Our work offers some methodological insights into shape-based analysis and aims to serve as a practical tool for the medical community. Code and data are openly available at: https://github.com/Francaexe/SSA-brainstem

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2841_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/Franca-exe/SSA-brainstem

Link to the Dataset(s)

The dataset is provided at following link under ‘data’ folder, containing a collection of original shapes of brainstem derived from both control subjects and pathological ones. We also provide meshes after preprocessing and remeshing procedure and the .mat file bringing from remeshed to original resolution. The link is the following: https://github.com/Franca-exe/SSA-brainstem

BibTex

@InProceedings{MacFra_Unraveling_MICCAI2025,
        author = { Maccarone, Francesca and Longari, Giorgio and Arrigoni, Filippo and Peruzzo, Denis and Melzi, Simone},
        title = { { Unraveling Brainstem Deformations in Joubert Syndrome: A Statistical Shape Analysis of MRI-Derived Structures } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15973},
        month = {September},
        page = {671 -- 681}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    In this paper, the authors extracted geometric features from MRI-derived brainstem data and utilized them to classify subjects with Joubert Syndrome (JS) versus healthy controls using a random forest classifier. The primary objective was to investigate the correlation between geometric deformations of the brainstem and the presence of JS, aiming to better understand the structural abnormalities associated with the syndrome.

  • 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.

    • Use of dataset-specific preprocessing: The authors applied preprocessing steps tailored to the characteristics of the dataset, which likely enhanced the quality and consistency of the input data for subsequent analysis.

    • Using different metrics for evaluation: The authors employed a comprehensive evaluation strategy using metrics like accuracy, specificity, sensitivity, and AUC. This multifaceted evaluation approach enhances the reliability of the proposed method’s performance assessment.

  • 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.

    • Lack of Citations in the Introduction The Introduction section would benefit from the inclusion of citations to support key claims and provide necessary context for the reader.

    • No Visual Examples of JS-Related Deformations Although the paper discusses MRI-based brainstem shape deformations caused by Joubert Syndrome (JS), no visual comparison is provided. Including representative examples of a JS patient’s brainstem versus a healthy control would significantly enhance clarity and support the clinical relevance of the work.

    • Missing Literature Review on JS Diagnosis The manuscript lacks a literature review on existing diagnostic methods for JS. A brief overview of current clinical and imaging-based approaches would help contextualize the novelty and relevance of the proposed method.

    • Missing Citation for Rigid Registration (Section 3.2) No citation is provided for the rigid registration approach. Adding a reference would support reproducibility and clarify the method used.

    • No Validation of the Statistical Shape Model (SSM) Standard validation metrics for SSM—such as compactness, specificity, and generality—are not reported. These are critical for assessing the quality and expressiveness of the model.

    • No Visualization of the SSM The paper would be strengthened by including visualizations showing how shape varies with the first few principal components (PCs). This would illustrate the model’s capacity to capture morphological variation.

    • Unclear Number of PCA Components Used for Classification The number of shape coefficients (principal components) retained for the classification task is not defined. This information is essential for understanding the dimensionality and complexity of the classification model.

    • Missing Discussion on Limitations of SSM A known limitation of SSM is that it captures only the variation present in the training data, which can restrict generalizability. This point should be discussed explicitly in the Discussion section.

    • Unexplained Abbreviations The acronym “KNN” is used without definition. All abbreviations should be spelled out at first mention to ensure clarity for a broad audience.

    • Insufficient Explanation of Figure 1(c) and 1(d) Subfigures 1(c) and 1(d) are not well explained in the text. Additional context should be provided to clarify their role and relevance.

    • Limited Reproducibility Due to Non-Public Data The study relies on an in-house dataset that is not publicly available, limiting reproducibility and the ability for others to verify or extend the findings.

    • Evaluation on a Single Dataset While initial evaluation on a single dataset is acceptable, testing on multiple datasets is generally recommended to assess the generalizability of the proposed method and reduce the risk of overfitting.

  • 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?

    The paper lacks clarity in several areas, with the most notable issue being the absence of validation for the generated Statistical Shape Model (SSM) after applying PCA. At a minimum, a compactness plot should be included to illustrate the amount of shape variation captured by each principal component. This is a standard practice in SSM studies and is essential for assessing the quality and representational capacity of the model.

  • 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.

    I recommend accepting this manuscript because it makes a novel, rigorously validated, and clinically meaningful contribution with only minor presentation tweaks remaining.



Review #2

  • Please describe the contribution of the paper

    The authors present a statistical shape analysis (SSA) method to identify pathological variations on 3D shapes extracted from MRI scans with a special focus on the correlation between the geometric deformations of the brainstem and Joubert syndrome (JS) pathology. To this end, the authors build upon existing literature and build a graph that encodes correspondences in terms of functional maps between each 3D shape in the collection. From which the authors can extract the deformation from each shape to a mean shape and build a PCA model creating a compact representation of the shape variations in the collection. The authors showcase the applicability of their method on a classification problem of JS. Finally, the authors offer a 2D/3D visualization tool.

  • 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.

    1- The proposed new initialization for shape collection matching is clearly presented and well-justified with quantitative evaluation (Table 1).

    2- The authors show a useful application of their SSA method on JS classification.

    3 - Overall, the paper is simple, clear and nicely written. The steps are simple and flow naturally one after the other.

  • 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.

    1- Methodological novelty is limited. The method builds mainly upon a similar workflow to the S4a method [16]. The main differences between the presented method and S4a are: a - The initialization step where the authors opted for a rigid alignment based on PCA computed on vertex locations while [16] used ICP to align each pair of shapes. b - The update step for up-sampling of the functional maps where the authors opted to up-sample the functional maps by one eigenfunction (similar to the original ZoomOut). So I think the difference between the proposed method and S4a is very limited.

    2 - Lack of comparison to any other method for the classification task. Multiple mesh-based SSA methods exist in the literature (list of methods contain but not limited to: FlowSSM (1), Mesh2SSM (2), S3M (3) …). With the current submission, I don’t see why the proposed method would be a better choice for this task compared to other methods.

    (1): Ludke et al. Landmark-free Statistical Shape Modeling via Neural Flow Deformations, MICCAI 2022 (2): Cates et al. Mesh2ssm: From surface meshes to statistical shape models of anatomy. MICCAI 2023 (3): Bastian et al. S3M: Scalable Statistical Shape Modeling through Unsupervised Correspondences. MICCAI 2023.

  • 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
    • Minor clarification: While choosing the mean shape, the paper states that the shape in the collection that is the “most similar” to the limit shape chosen. It is not clear what is meant by “most similar”, a clarification would be beneficial to the clarity of section 3.
  • 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?

    The main factors contributing to my decision are, first, the lack of comparison to any other method in the literature and, second, the limited methodological novelty. (see my explanation in weaknesses).

  • Reviewer confidence

    Confident but not absolutely certain (3)

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

    Reject

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

    I am leaning towards rejecting this paper since the authors didn’t clarify their contribution compared to the S4a method which was my main concern. Additionally, the lack of comparison to other methods is a bit concerning. Some of the methods I listed in my review are publicly available and do share their own datasets as well. The authors claimed in their rebuttal that they cannot compare their methods to competitors which I don’t completely agree with. For these reasons, I opt for rejecting the proposed work.



Review #3

  • Please describe the contribution of the paper

    The paper uses existing techniques albeit on a new application. The new application is the statistical shape analysis of brainstems of Joubert Syndrome (JS) patients. The techniques employed include: Segmentation using FreeSurfer; Meshing using 3D-Slicer; Rigid registration by aligning the first PC; Mesh ReMatching to reduce density and noise; Initialization of vertex correspondence using nearest neighbor (vertex); (group) shape correspondence using the functional maps framework (FMAP) and ZoomOut algorithm. The authors customize the computational pipeline by making specific choices of settings and parameters to fit their (SSA of Brainstem in JS) application. The authors use the PCA coefficients and the rotation correction angle to train a random forest classifier. The rotation angle is used since, as the authors state, “this rotation angle represents an important feature that highly correlates with the occurrence of the pathology”.

  • 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 paper appears sound and is largely well written.
    • The work leverages a sequence of existing techniques for a new anatomical statistical shape analysis application.
    • The work uses PCA alignment to initialize the shape correspondence (instead of the standard FMAP initialization, which cut the computation time for this initialization by about half while yielding similar alignment (assessed via the Chamfer distance).
  • 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.
    • Restricted shape analysis.
    • Lack of clarity on how some of the steps are performed.
  • Please rate the clarity and organization of this paper

    Good

  • 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 submission does not provide sufficient information for reproducibility.

  • 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

    It is surprising that the authors did only used PC coefficient and the angle, ignoring the shape (deformation) features when analyzing differences between brainstem structure in controls vs JS patients. Nevertheless, they achieved classification accuracy, precision, recall, and F1 scores around 80-90%. The authors write “RF classifier correctly identifies JS patients based on the SSA-derived features”, but it unclear what the SSA features are (aside from the PCA coefficients and the correction angle).

    The paper then examine the deformations between shapes (presumably as dictated by the vertex correspondence resulting from FMAP), and visualizes the compression and “dilation” (perhaps via determinant of Jacobian calculation, but that’s not explicitly stated, and the cited 1999 paper [4] does not show similar visualizations).

    This sentence is unclearly written. “FMAPs exploits the fact that a given vertex-wise map Π12 uniquely induces a map T21 : F(S2) →F(S1), from which it is then possible to recover an approximation of Π12.”

    The “nearest neighbor search in the 3D space” is not clearly described.

  • 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?

    A new application (JS disease) of shape correspondence using an existing correspondence methodology that is not typically presented at MICCAI. Leverages multiple computational methods to serve the application. Well written paper. Shows effective correspondence initialization (speed and accuracy).

  • 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.

    My ‘weak accept’ persists. I believe the paper will even be improved after addressing all reviewers concerns in the final revised version of the paper.



Author Feedback

We thank the reviewers for their valuable feedback. As recognized by R1 and R2, our pipeline is an application study extending [16] for the specific diagnosis of JS. Moreover, focusing on children with a rare disease, dataset availability is limited, disarming learning-based solutions. In this context, we prioritize SSA over SSM to demonstrate that our geometric analysis provides a valuable diagnostic pipeline, while leaving modeling to further explorations. To our knowledge, this is the first SSA approach applied to JS, paving the way for broader application to other datasets and pathologies.

Limited Comparisons and Methodological Novelty (R2,R3) While we stress again that we consider our work an application study, we agree with R3 that some state-of-the-art methods can be considered competitors. We tested learning-based alternatives like Mesh2SSM (2) but found them ~24× slower and less effective on small datasets, consistent with closed issue#3 on the Mesh2SSM GitHub project, highlighting its dependence on larger data. These issues make these solutions unsuited for our work, which aims to provide an efficient pipeline to address a rare disease (JS) with limited available data. Scarce availability of data also bounds possible evaluations. We are open to testing on external benchmarks, but none were available during our study. In this light, the release of our dataset, our code and our 2D/3D visualization tool are novel contributions that foster clinical and diagnostic translation.

Missing Visualizations (R2) Due to space limits and the current policy, which prohibits figures in the supplementary materials, we prioritize figures showing the efficacy of SSA in characterizing JS outcomes. For these reasons, we excluded images we prepared for template variations along PCs, PCs cumulative variance (PCs images), and examples of JS-related deformations. At the same time, we depicted the shape on the right of Fig.2 (Sec.2.3) with exaggerated deformations to emphasize common pathological changes. We will refer to [22] for visual insights of JS structural deformations and add a close-up of an axial slice showing MTS in a JS patient by reorganizing Fig.2. If the reviewers agree, by referring to the released code for the data preparation details, we will save enough space to include the PCs images that also address the next point. Moreover, we will revise figures and captions to integrate all the suggestions.

SSM and PCA (R1,R2) While our focus was on SSA, we agree on the value of SSM metrics that we could explicitly add. However, we assess them indirectly: For generality, we reported 8-fold cross-validation, consistently achieving strong classification scores. For specificity, the new image on the template variations along PCs will show the generation of plausible artificial shapes. For compactness, although all PCs and rotation angles were input to the RF, only the first 10 significantly contributed to performance. We acknowledge the lack of detail on PCA. In the CR, we will clarify that the first 10 PCs explain ~80% of the variability and mainly encode localized deformations (e.g. top/bottom or asymmetries).

Missing Context (R2) We agree to anticipate key citations in Sec.1 for better contextualization. Furthermore, we will highlight that JS diagnosis relies on visual detection of the MTS during reporting. Artifacts in the image can lead to incorrect or inconclusive assessment, emphasizing the importance of the proposed 3D approach (Sec.2.3).

Reproducibility (R1,R2) As mentioned at the end of Sec.1, upon acceptance, we will release our dataset and code to ensure full reproducibility and facilitate application to other datasets. Our implementation includes the novel rigid registration solution we tailored for the specific shapes involved, for which we cannot provide any reference.

Minor comments (R1,R2,R3) We will incorporate all the minor suggestions highlighted in the reviews, discuss our limitations, and clarify the acronyms.



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.

    Reject

  • 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’

    This paper presents a statistical shape analysis pipeline tailored for identifying brainstem variations in Joubert Syndrome patients. Reviewers #1 and #2 highlighted the clinical importance and clarity of the application, while Reviewer #3 raised concerns about limited methodological novelty and comparative analysis. The authors’ rebuttal effectively addressed these concerns, emphasizing the paper’s application-oriented nature and constraints due to rare disease datasets. Given its clear clinical value and sufficient validation, the paper warrants acceptance. Remaining comparative studies can be pursued in future work.



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’

    The paper primarily employs existing methods to analyze shape variations in brainstem deformation associated with JS. Despite relying on established techniques, it presents a novel application to a rare disease and demonstrates promising results.



back to top