List of Papers Browse by Subject Areas Author List
Abstract
Tau pathology is a hallmark of Alzheimer’s disease (AD), and longitudinal tau positron emission tomography (PET) provides valuable insights into disease progression. However, the integration of tau PET data into computational models remains limited by challenges in encoding topographical information and ensuring longitudinal consistency. Existing biomarker-based representations often lack spatial flexibility and fail to account for covariance between brain regions. Additionally, traditional approaches often treat longitudinal scans as independent observations, neglecting temporal coherence. To address these limitations, we propose a novel Multiresolutional Reeb Graph representation that encodes the spatiotemporal propagation of tau topographical information. Our method constructs Reeb graphs to capture tau topography at a static time point and extends them into a multiresolutional framework to model disease evolution. We introduce a topology-based measurement for quantifying pathology spatial distribution similarity, and a severity interleaving distance for robust longitudinal staging. The efficiency of the proposed representation is validated in two downstream tasks: an integrated subtyping and staging system, and the longitudinal pathology prediction. The promising results compared with the current methods demonstrate the great potential of the proposed representation to enhancing the application of longitudinal tau PET data, and offering a reliable approach for studying AD progression.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/1465_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{YueJia_Robust_MICCAI2025,
author = { Yue, Jiaxin and Zhang, Jianwei and Wang, Xinkai and Shi, Yonggang},
title = { { Robust Topographical Representation for Longitudinal Propagation of Tau Pathology } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15971},
month = {September},
page = {583 -- 592}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper uses Reeb graphs to capture the topographical information from tau-PET data. The advantages of this approach are that it can capture covariance between brain regions and can incorporate temporal coherence for longitudinal modelling. The authors present metrics for assessing the topographical similarity of tau-PET measurements as well as their relative disease severity. This facilitates modelling of the spatiotemporal progression of tau using multi-resolutional Reeb graphs. They use this framework to subtype and stage tau-PET data from ADNI, and for a prediction task using longitudinal data.
- 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 work addresses an important clinical application: tau-PET is an emerging biomarker for AD progression, so novel methods for longitudinal modelling and prediction have clinical relevance. The application of topological modelling is well motivated and the directed graph approach for subtyping and staging is interesting. The results of the subtyping and staging experiment are benchmarked against the popular SuStaIn algorithm.
- 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.
My main concern relates to the novelty of the proposed approach. A previous MICCAI paper, by Yue et al. (2023), demonstrates a very similar application of Reeb graphs to tau-PET data: https://pmc.ncbi.nlm.nih.gov/articles/PMC10951551. In this paper they also use Louvain community detection for subtyping and staging and compare with the SuStain algorithm. It is not clear what are the advantages of the proposed approach compared to the approach presented by Yue et al.
Furthermore, there are some methodological details missing from the experiments section, eg. the number of nearest neighbours K for the longitudinal prediction task, and the regularisation parameter, theta. A cross-validation would also help to assess the stability of the proposed method.
- 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
For future work, it would be interesting to see a more comprehensive comparison of the subtypes and stages derived from this method with those from SuStain, since the current manuscript only addresses SuStain’s longitudinal stability. It would also strengthen the manuscript if the subtypes from the Reeb graph method could be validated in a different dataset.
The Reeb graph framework is used for a longitudinal prediction task, where the SUVR pattern at time t-1 is used to predict the subsequent pattern at time t (I think, it’s not completely clear in the manuscript how many preliminary scans are used to make the prediction). If this is the case, the authors achieve superior longitudinal prediction accuracy from only one scan, compared to the other methods which require two scans. I think this is a key strength of the method compared to others and should be highlighted.
There are some typos, eg section 3.2: “our proposed longitudinal prediction approach reach the promising accuracy, outperforming than other methods.” -> “our proposed longitudinal prediction approach achieves promising accuracy, outperforming other methods.”
- 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.
(2) Reject — should be rejected, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Unfortunately the project is very similar to a previous work presented at MICCAI, which means that it is not sufficiently novel to be accepted.
- Reviewer confidence
Somewhat confident (2)
- [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.
Thank you to the authors for the rebuttal. Although I am pleased that they will now add in the reference for the Yue et al. paper, I still think that the works are still too similar and I recommend the paper to be rejected due to lack of novelty.
Review #2
- Please describe the contribution of the paper
The paper uses a reeb graph-based method to characterize the coarsened spatial distribution of tau PET images. It creates multi-resolution Reeb graphs through gradual node merging. It also introduces a novel approach for subtyping and staging following the reeb graph. The authors validate the model performance using a longitudinal prediction scheme.
In the experiments, the authors validate their method on 368 ADNI subjects. The results demonstrate good prediction capability, and the subtyping approach reveals differences in demography, age, and MMSE drop rate.
- 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.
I. The usage of reeb graph and further multiresolution reeb graph is a novel approach. It can help measure similarity of the image at various levels of resolution. Also, it can help longitudinal prediction of tau, which is important in the field.
II. Many parts of the method are demonstrated in detail.
III. Part of the results echo the aims of the paper, especially Fig. 2.
- 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.
There are a few downsides of the paper: I. The writing in section 2.1 can be more clear. The author should illustrate clearly how they compute node value and edge value of the reeb graph clearly in math formulation. In section 2.3 “Subtyping and Staging” part, the writing is also unclear. How do you define the Q, c_x and q_x? What are the unknown variables to be optimized in equation 5 and equation 6?
II. I can not see the purpose of Fig. 3 (b)(c). If we split the patients into three clusters arbitrarily, they will also demonstrate differences between groups. What are the insights of Fig. 3(b)(c). For Fig. 3(d), Are subtype 2 and 3 similar in the MMSE drop rate?
III. What is the level of reeb graph when you compare results in Table 1?
- 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 mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure 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
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.
(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?
The proposed multiresolution reeb graph model is novel in this field. In the preliminary results shown in the paper, it demonstrates promising tau-PET prediction ability.
The paper also has some unsatisfactory perspectives. For example, the section 2.1 and 2.3 should be written more clear. The results in Fig. 3 should discuss the clinical insights instead of just showing differences.
- 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 author specifically answered the questions I mentioned in my review. I think the paper’s idea is justified by the response.
Review #3
- Please describe the contribution of the paper
Reeb graph analysis was used in this work to characterize the spatial distribution of tau pathology as indicated by PET imaging. Multiresolutional Reeb graphs were then used to encode temporal propagation of tau. This reperesentation was then subtyped (for topography of tau uptake patterns) and staged (to indicate progression), with the ability to predict/interpolate/extrapolate SUVRs.
- 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.
Novelty: using MRG for tau analysis and with it the ability to predict tau SUVRs in the whole image as the tau pathology progresses.
- 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.
Though the subtyping and staging is novel, how this method compares/contrasts with current staging of tau (such as Braak staging [4]) and thus whether this method supports current knowledge or not would be of interest. The validation of interpolation/extrapolation relies on a single metric, the MSE, on the whole image, which would not be representative enough of the accuracy of the tau PET prediction, as tau uptake might be focal.
- 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 submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure 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
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.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The novel use of MRGs as well as tau-related research (generally fewer in comparison to other widely-used radiotracers) would be of interest to the audience of this conference.
- 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.
Most of this reviewer’s comments have been addressed adequately. The comment on the focal uptake of tau imaging has been recognized but could still be further discussed (barring extra experiments in a future work) to improve the impact of the work.
Author Feedback
First, we would like to thank all reviewers for your time and valuable comments on our work. Following the suggestions of reviewers, we respond below to reviewers’ comments.
- Novelty and advantage over Yue et al (Reviewer 2). While our work and Yue et al. both utilize the foundational tools including Reeb graph and persistence, they differ significantly in purpose and technical contribution: (1) Different problem setting: Yue et al. focuses on subtyping from cross-sectional data. In contrast, our work addresses longitudinal modeling using a robust multi-resolution representation. (2) New spatiotemporal metrics: We introduce novel similarity measures (Eq. 2–4) designed specifically for longitudinal data. (3) Modified graph analysis: Instead of applying standard community detection, we adapt the graph structure and algorithm for sequential staging and temporal subtyping. (4) While subtyping is included as a downstream task, the core strength of our framework lies in its ability to model multi-time-point data—both in the design of the representation and in the directed graph structure. We will revise our manuscript to add a proper citation and discussion of the above points. 2.Clinical insights (Reviewer 1, 3). Subtype 1(Fig. 3(a)) exhibits patterns consistent with Braak stages and align well with established knowledge in pathology. The purpose of Fig. 3(b)(c) is to illustrate the distinct distributions and characteristics of each subtype. In Fig. 3(b), each subtype includes subjects across all diagnosis groups, indicating that the observed differences are not due to data imbalance. Additionally, our preliminary subtyping reveals significant differences in age distributions (p < 0.05; Fig. 3(c)), which could not be guaranteed by random grouping. While Subtypes 2 and 3 show similar overall MMSE drop rates (Fig. 3(d)), their stage-specific differences suggest variable disease propagation speeds. This underscores the need for a more precise subtyping and staging model. We acknowledge that the current clinical exploration is preliminary and plan to investigate it further in future studies.
- Validation of longitudinal predictions on focal patterns (Reviewer 1). In this study, we evaluate prediction accuracy using MSE as a global measure. We acknowledge the importance of analyzing focal patterns and plan to apply our Reeb graph tool to explore these localized changes in future work.
- Clarification of longitudinal prediction (Reviewer 2). We confirm that our method requires only one previous scan to make the longitudinal prediction, and this could achieve better results over other interpolation or extrapolation methods that require at least 2 scans.
- Parameter selections (Reviewer 2). We set K=5 and theta=1e-3 based on preliminary experiments on a subset of the data. We will use cross validation for selecting the optimal parameters in our future work as suggested.
- Clarification of MRG construction, subtyping and staging (Reviewer 3). For section 2.1, node values in the Reeb graph represent SUVR at critical points, while edges indicate the connectivity between these points and do not carry values. For section 2.3, Q is the modularity of the graph, and its formal definition is in reference [2]. c_x refers to the subtype membership of the scan, which is randomly initialized following [2] and optimized during training. q_x refers to the stage membership of the scan and is also optimized by the algorithm. We will revise these sections in the manuscript to improve clarity.
- Level of Reeb graph (Reviewer 3). We take 6-level multi-resolution Reeb graph in Table 1 as this level of complexity could sufficiently capture the majority of spatiotemporal features. For future work, more experiments can be conducted to explore the optimal choice of this parameter.
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’
N/A
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’
accepts