List of Papers Browse by Subject Areas Author List
Abstract
Alzheimer’s Disease (AD) and Lewy Body Dementia (LBD) often exhibit overlapping pathologies, leading to common symptoms that make diagnosis challenging and protracted in clinical settings. While many studies achieve promising accuracy in identifying AD and LBD at earlier stages, they often focus on discrete classification rather than capturing the gradual nature of disease progression. Since dementia develops progressively, understanding the continuous trajectory of dementia is crucial, as it allows us to uncover hidden patterns in cognitive decline and provides critical insights into the underlying mechanisms of disease progression. To address this gap, we propose a novel multi-scale learning framework that leverages hierarchical anatomical features to model the continuous relationships across various neurodegenerative conditions, including Mild Cognitive Impairment, AD, and LBD. Our approach employs the proposed hierarchical graph embedding fusion technique, integrating anatomical features, cortical folding patterns, and structural connectivity at multiple scales. This integration captures both fine-grained and coarse anatomical details, enabling the identification of subtle patterns that enhance differentiation between dementia types. Additionally, our framework projects each subject onto continuous tree structures, providing intuitive visualizations of disease trajectories and offering a more interpretable way to track cognitive decline. To validate our approach, we conduct extensive experiments on our in-house dataset of 308 subjects spanning multiple groups. Our results demonstrate that the proposed tree-based model effectively represents dementia progression, achieves promising performance in intricate classification task of AD and LBD, and highlights discriminative brain regions that contribute to the differentiation between dementia types. Our code is available at https://github.com/tongchen2010/haff.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/3485_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/tongchen2010/haff
Link to the Dataset(s)
N/A
BibTex
@InProceedings{CheTon_AUnified_MICCAI2025,
author = { Chen, Tong and Chen, Minheng and Zhang, Jing and Zhuang, Yan and Cao, Chao and Yu, Xiaowei and Lyu, Yanjun and Zhang, Lu and Su, Li and Liu, Tianming and Zhu, Dajiang},
title = { { A Unified Continuous Staging Framework for Alzheimer’s Disease and Lewy Body Dementia via Hierarchical Anatomical Features } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15962},
month = {September},
page = {12 -- 22}
}
Reviews
Review #1
- Please describe the contribution of the paper
- A multi-scale feature fusion framework is proposed to capture hierarchical anatomical features and model the continuous transitions across different stages of dementia.
- This study employs a graph neural network-based model to integrate structural connectivity (SC) and cortical folding pattern features at different scales. It captures the fine-grained regional differences that can able to differentiate AD and LBD.
- 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 study proposes a novel multi-scale learning framework that utilizes hierarchical anatomical features to model the continuous relationships among various neurodegenerative diseases, including mild cognitive impairment (MCI), Alzheimer’s disease (AD), and Lewy body dementia (LBD). By projecting each subject onto a continuous tree-like structure, the framework intuitively visualizes disease trajectories, offering a more comprehensible approach to the study of neurodegenerative disorders.
- 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.The study lacks detailed descriptions of the model architecture, such as the number of GCN layers and Transformer heads, as well as the hyperparameter tuning process. Furthermore, the rationale behind selecting the 3-Hinge Gyrus (3HG) instead of other cortical folding features remains unexplained. 2.In Fig. 1, the Multi-scale Graph Embedding Fusion (MGEF) framework includes four EPs, but the sources of these EPs are unclear. Additionally, it remains unspecified whether Eq is used as an input. 3.In Table 1, does the F1 score for the “Our method” row lack standard deviation? Moreover, the compared methods listed in Table 1 are outdated and should be contrasted with state-of-the-art (SOTR) methods from the past two years. 4.The study relies solely on an internal dataset (308 cases, 77 cases per group), which is relatively small. The absence of validation on public datasets (e.g., ADNI) may limit the model’s generalizability. 5.The 3-Hinge Gyrus (3HG) was selected as a key feature, but it has not been compared with other cortical folding patterns (e.g., 2-Hinge, 4-Hinge) or traditional gyral-sulcal metrics. 6.It is recommended to include an analysis of the contributions of features across different scales, such as large-scale atlas regions and small-scale 3-Hinge Gyrus (3HG).
- 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
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 manuscript presents a multi-scale framework for modeling dementia progression but faces three critical limitations: (1) Limited generalizability due to a small in-house dataset (n=308) and absence of external validation on public cohorts (e.g., ADNI); (2) Suboptimal classification performance (61.68% accuracy for four-class task) with no mitigation of class imbalance or use of robust metrics (e.g., AUROC); and (3) Methodological gaps, including instability in disease progression visualization (Fig. 3), insufficient technical details (e.g., GCN/Transformer configurations), and lack of biological interpretation linking discriminative regions to AD/LBD pathology. Addressing these issues is essential to strengthen the framework’s clinical relevance and technical rigor.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
The authors proposed a novel multi-scale learning framework that leverages hierarchical anatomical features to model the continuous relationships across various neurodegenerative conditions, including Mild Cognitive Impairment, AD, and LBD. Their approach employ the proposed hierarchical graph embedding fusion technique, integrating anatomical features, cortical folding patterns, and structural connectivity at multiple scales. This integration captures both fine-grained and coarse anatomical details, enabling the identification of subtle patterns that enhance differentiation between dementia types.
- 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 employs a graph neural network-based model to integrate structural connectivity(SC) and cortical folding pattern features at different scales so that it captures the fine-grained regional differences that’s able to differentiate AD and LBD.
-
It constructs a tree structure for all subjects, easier to understand the relationship between similar diseases, such as AD and LBD.
-
It used 3-hinge Gyrus, which is shown to be more predictive of AD stages, and generates relatively good results.
-
- 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.
- When comparing CN, MCI, AD and LBD, the accuracy is not so good, and is only compared with machine learning methods, but not available deep learning methods.
- 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 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.
(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?
I like the tree-based classification graph. It allows us to understand the relationship between different neurodegenerative diseases.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
The main contribution of this paper is a multi-scale learning framework that models the continuous disease progression between CN, MCI, AD and LBD. The authors leverage a hierarchical graph embedding fusion framework to integrate anatomical features derived from atlas-based regions and 3-Hinge gyrus hubs for structural connectivity. By projecting subjects onto continuous tree structures, the approach provides interpretable visualizations of disease trajectories.
- 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 authors propose a novel methodological framework based on existing techniques such as GCN, tree learning and 3HG.
- The use hubs defined based on 3-Hinge gyrus cortical folding patterns in addition to solely atlas-based regions is interesting
- The authors perform 5-fold cross validation using an in-house dataset, with a balanced distribution of diagnoses.
- 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.
Based on the comparisons shown in Table 1, it is not clear whether the improved classification performance is due to the method or use of both cortical folding features and standard SC features. The authors do not include baseline comparisons (like Random Forest or SVM) that also leverage this combined CF+SC feature set, making it impossible to isolate the contribution of their framework.
- The result of the framework depends on the learnable prototypes which seem to vary significantly (Fig 3).
- 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
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 authors present a novel methodological framework that integrates GCNs, tree learning, and novel 3-Hinge gyrus-defined hubs alongside traditional atlas-based regions to classify disease stages. Furthermore, the experiments are well-designed on a well-balanced dataset.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- 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