List of Papers Browse by Subject Areas Author List
Abstract
Distinguishing progressive mild cognitive impairment (pMCI) from stable MCI (sMCI) is crucial for timely treatment of Alzheimer’s disease (AD), yet it is challenging due to inherent class imbalance and limited data. While recent data synthesis methods have shown successful results, they often disregard distributional differences between groups and individual heterogeneity in disease progression. Also, they treat the whole-brain as a unified entity, overlooking region-specific features despite their varying associations with AD. To address this, we propose a novel end-to-end framework that augments MCI data and predicts their future conversion to AD. This is realized by using adversarial attacks that directly control data points in the feature space considering group differences. The attacks are adaptively applied with region-wise learnable attack intensities and subject-specific attack steps, which are flexibly adjusted based on each subject’s observation interval. Moreover, we introduce a trajectory constraint that ensures the attacked (i.e., augmented) data follow plausible disease progressions and preserve realistic neurodegeneration patterns. Extensive validation on two AD biomarkers across three classifiers shows our method’s superiority over six baselines.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2850_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{ChoHyu_Adaptive_MICCAI2025,
author = { Cho, Hyuna and Ahn, Hayoung and Wu, Guorong and Kim, Won Hwa},
title = { { Adaptive Adversarial Data Augmentation with Trajectory Constraint for Alzheimer’s Disease Conversion Prediction } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15966},
month = {September},
page = {13 -- 23}
}
Reviews
Review #1
- Please describe the contribution of the paper
1.The authors propose a new framework for augmenting pMCI samples and predicting their AD conversion.
2.The authors develop the trajectory consistency regularization to ensure that the augmented samples follow plausible disease progression.
- 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.
-
It is good to see that the proposed method has been compared with a wide range of existing data augmentation and sample synthesis approaches.
-
It is good to see that the authors adopt multiple classififers to validate the effectiveness of the proposed data augmentation method.
-
The authors have reported the mean and standard deviation for all the numerical 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.
I only have two minor concerns that needs to be addressed by the authors.
- The latest work cited by this paper, is published in 2021 (4 years ago). It is strongly suggested to cite more new papers and discuss the difference between them and the proposed method. For example, it is suggested to cite VAPL[a] and UMML[b].
[a] Kang, Luoyao, et al. “Visual-attribute prompt learning for progressive mild cognitive impairment prediction.” International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer Nature Switzerland, 2023. [b] Feng, Yidan, et al. “Unified Multi-modal Learning for Any Modality Combinations in Alzheimer’s Disease Diagnosis.” International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer Nature Switzerland, 2024.
- Will the authors release the source code (including training, inference code and trained model weights)? It is suggested to release the source code.
- 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.
(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?
It is a very interesting work with novel techinical contributions. The authors also provided extensive and convinced results to show the strength of the proposed method.
- Reviewer confidence
Very confident (4)
- [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
Novel end-to-end framework for synthesizing small-size pMCI data and predicting their AD conversion.
- 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 did a strong evaluation of their method on two different datasets. As well they chose good baselines methods.
- 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.
n/a
- 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?
Extensive experiments across multiple datasets and classifiers validate the effectiveness of their proposed method.
- 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
This paper presents a novel data augmentation method for synthesizing progressive Mild Cognitive Impairment (pMCI) samples using adversarial attacks. Specifically, the authors perturb data in the feature space to simulate progression along the Alzheimer’s disease trajectory. The attack steps are adaptively scaled based on the subject’s observation interval, and the perturbation magnitudes are learned in a region-of-interest (ROI)-wise manner to reflect dynamic changes in specific brain regions. A trajectory consistency constraint is also introduced to ensure the biological plausibility of the generated samples.
- 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 is well-written and clearly motivated
The proposed method is novel and addresses a meaningful challenge: the imbalance of MCI samples. The authors try to model the progression of pMCI subjects using longitudinal datasets and perturb the real pMCI data along the progression trajectory to generate new pMCI samples.
- 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.
See comments below
- 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
- It appears the classifier is not pretrained but jointly optimized with the adversarial attack. If this is the case, during early training, the classifier is random and its gradients may be unreliable. This could lead to unstable or random adversarial updates in Equation (1). How do the authors ensure stable convergence in this situation?
- The number of adversarial attack steps is defined as a linear function of the time interval (i.e., M/2). However, Alzheimer’s progression is often nonlinear — some subjects may show minimal early changes but decline rapidly later. A linear assumption might overlook such dynamic.
- 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 paper introduces novelty by using adaptive attack steps, ROI-wise magnitude for adversarial attacks. The authors also use a consistency constraint to ensure that the generated samples roughly fit the real disease progression trajectory. Furthermore, the experiments are thorough and prove the effectiveness of proposed components.
- Reviewer confidence
Somewhat confident (2)
- [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
We thank all reviewers for their unanimous acceptance and constructive reviews. Here we clarify all questions and concerns.
Q) Rev #1: Add more baseline methods. A) We thank the reviewers for these suggestions. We will include these baselines and compare them with our method in detail in the future journal version of our paper.
Q) Rev #1: Will the authors release the code? A) Yes, we will publicly release the source code and the trained model weights on both datasets.
Q) Rev #3: How to stabilize convergence in the early stages of training? A) As the reviewer correctly understood, the classifier is not pretrained on the given data; instead, it is jointly trained from scratch along with adversarial magnitudes on both the original and augmented data. The motivation for adopting this end-to-end scheme (i.e., from data augmentation to classification) is to prevent potential biases and overfitting that could arise from training only on the highly imbalanced given data.
However, as the reviewer pointed out, this setup can lead to unstable adversarial updates during the early training. To address this, we slowly decreased learning rates over epochs using a learning rate scheduler, which helps smooth out early training instabilities. With the scheduler, we empirically observed that the model eventually converged reliably over the training process.
Q) Rev #3: A linear assumption of disease progression may overlook the nonlinear progression of AD. A) We thank the reviewer’s comment and agree with this opinion. However, capturing the true dynamics of complex, nonlinear longitudinal disease progression requires a much larger number of sample points collected at finer temporal resolution than is currently available in our dataset. Without these dense ground truth data, assuming reliable non-linear disease dynamics is very challenging and may lead to misinterpretation of temporal patterns. In future work, we will search for longitudinal samples with more follow-up time points and explore advanced modeling schemes to better capture the true disease progression.
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