List of Papers Browse by Subject Areas Author List
Abstract
Monitoring progression from Mild Cognitive Impairment due to Alzheimer’s Disease (MCI-AD) is critical for patient care. However, current approaches to model AD progression overlook complex interrelated neurodegeneration in different regions of the brain and how AD pathology and genotypes manipulate it. This study defines neurodegeneration dynamics and proposes the Dynamics Individualized by Static Covariates without LOngitudinal ScrEening (DISCLOSE) framework. This method predicts individualized neurodegeneration dynamics from only baseline amyloid-beta deposition and the number of APOE4 alleles with an Ordinal Differential Equation (ODE). We evaluated DISCLOSE using longitudinal MRI samples in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. The results demonstrate that DISCLOSE outperforms existing methods in long-term trajectory prediction, particularly for predictions beyond three years. This work presents a significant step toward modeling individualized disease trajectories. Also, DISCLOSE could quantitatively interpret the effects of AD-related genotypes and pathophysiology on regional atrophy progression.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/1182_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{JunWoo_DISCLOSE_MICCAI2025,
author = { Jung, Wooseok and Park, Joonhyuk and Kim, Won Hwa},
title = { { DISCLOSE the Neurodegeneration Dynamics: Individualized ODE Discovery for Alzheimer’s Disease Precision Medicine } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15974},
month = {September},
page = {178 -- 187}
}
Reviews
Review #1
- Please describe the contribution of the paper
The authors introduce DISCLOSE, a novel framework designed to predict individualized brain atrophy dynamics using only baseline features. This enables subject-specific simulation and longitudinal monitoring without requiring longitudinal data. Through this biologically informed personalization, DISCLOSE demonstrates improved predictive performance in modeling brain atrophy 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.
1) The proposed framework achieves high-fidelity individualized predictions using only baseline measurements, thereby alleviating the burden of longitudinal data collection and making it more practical for clinical deployment. 2) The explicit incorporation of APOE4 genotype and amyloid burden improves the biological plausibility of the model and enhances its ability to capture individual variability in neurodegeneration trajectories.
- 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) Although the authors acknowledge that the proposed model lacks explicit sparsity and is limited in handling inherent data variability, it remains unclear what methodological contribution the model makes. Based on the results reported in Table 1, while the authors argue that DISCLOSE consistently outperforms baseline methods across different time points, the prediction error at earlier time points ($\tau < 1$) appears relatively high and may require further discussion regarding its clinical significance. It would be helpful if the authors could provide a more in-depth analysis of why the model performs poorly at these early time points. 2) In addition to the limitations regarding the generalizability of the proposed model, it would be beneficial for the authors to compare their in-house preprocessing pipeline with more commonly used, publicly available methods. Such a comparison would help assess the reliability and reproducibility of the data processing steps and strengthen the validity of the overall framework.
- 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 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
[Minor] 1) In the related work section, it is unclear whether the cited study [5] is directly relevant to the “MCI to Dementia Conversion Prediction” subsection. 2) While the model’s focus on APOE4 and amyloid burden as primary covariates for neurodegenerative dynamics is reasonable, the authors are encouraged to add references to substantiate the biological rationale behind this assumption. 3) Figure 1 should include the full name of the acronym “CN” for clarity. 4) Additionally, as “amyloid SUVR” is introduced earlier in the manuscript (Section 3.1 Study Dataset), its full name should be provided upon first mention for clarity. 5) To improve clarity and consistency, the numerical precision in Table 1 should be standardized (e.g., L1+diff, τ > 3: 0.06 should be formatted as 0.060).
- 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?
see the comments
- 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
Presents a novel method to fit a patient specific curve for longitudinal tracking of brain atrophy considering a population based curve.
- 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.
-Well written (aside from a few points below) with a clear state of the art -Good validation of the methodology including an ablation study + error bounds on the curve fitting
- 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 ablation study demonstrates that the Loss function with 3 terms is probably not the best design – both L1 only and L1+diff are better for some measures and almost always within the standard deviation for others (except SP). What justification is there to use the more complex loss function given the results? -In my opinion leading with the mathematical theory without describing why it is relevant for this problem and how it is informing the solution derived is confusing. It is also not clear why Picard’s theorem is so essential. I would first start by describing the problem and the aim and then introduce only the mathematics necessary to explain the approach taken. -Some acronyms (IVP, SINDy) are used without being defined first -It is not clear how INSITE and SINDy are related, especially when comparing against SINDy (table 1) how is this different from INSITE? -some variables (epsilon) are used twice to indicate different concepts -one equation in Remark is unnumbered
- 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.
(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?
Overall I think this is an interesting paper with a good methodological contribution. However I have some concerns over the final loss function given their results that I would like the authors to address and given their response could change my opinion of the paper. There are also some other minor points I would like clarified.
- 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.
Overall, I think this manuscript is borderline. However, I think most of the reviewer negative points is more to do with -the lack of clarity of the paper, -context of other work in this field, focusing on theory rather than details of the experimental design and results.
I think these are all perfectly addressable by the authors when preparing the camera ready version. As long as they revise keeping in mind much of the MICCAI community will not have subject matter expertise in modeling temporal dynamics/disease progression and such more explanation for some details needs to be provided I think this will be suitable for publication.
Review #3
- Please describe the contribution of the paper
The authors used APOE4 allele status and 18F-AV45 amyloid PET SUVR to predict longitudinal volumetric changes in segmented brain regions as measured by T1 weighted anatomical MRI.
- 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 model appears to successfully predict brain atrophy in the ADNI dataset. It will be interesting to see how well the model generalizes to prospectively collected data.
One of the focuses the results and discussion is the connectivity of different brain regions, which appears to be an output of the DISCLOSE model. These connectivity values would be interesting to compare to connectivity as measured by fMRI. It may imply that fMRI connectivity could be a valuable input for the model. Relatedly, it would be interesting to see if the predicative power of fMRI connectivity is correlated with the covariates already included the model (APOE4 and amyloid burden).
- 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 implications of the work should be explored more in the discussion. The presumed utility of such a method, which is referenced in the introduction, would be to help demonstrate the efficacy of a novel treatment for AD, particularly if the treatment effect is confounded by APOE4 status or existing amyloid burden.
Relatedly, the time buckets, tau, appear to be based on data availability rather than chosen to evaluate the effectiveness of a proposed intervention. Given the dearth of options for treating AD, predicting neurodegeneration at < 1 year, may be only so helpful, and predicting at 3 years onward may be too wide a time bin.
I found the paper dense and difficult to read.
- 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 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?
Overall, the work is detailed and thorough, though equivalently dense and difficult to follow. More discussion of the clinical implications would be appreciated.
- 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.
No new notes. I’m satisfied with the authors’ rebuttal.
Author Feedback
We appreciate the reviewers’ thoughtful feedback on our manuscript and the opportunity to address their concerns.
Lacking Readability (R1 – R3) (R1) In the introduction, we will clearly rationalize amyloid-beta plaques and APOE alleles as AD biomarkers and risk factors, respectively. (R2) Picard’s theorem was introduced to emphasize Lipschitz-ness for existence and uniqueness of an ODE (in our case, f(x) is linear hence Lipschitz). We aimed to illustrate what DISCLOSE is and some underlying mathematics initially, then move on problem definition (an introduction to neurodegeneration dynamics), and finally explain how DISCLOSE works with the problem. Still, we agree that referencing the whole theorem is unnecessary. (R3) The neurodegeneration dynamics allows us know how different brain regions are intercorrelated and can predict the onset time of severe neurodegeneration. This will potentially support triaging high-risk MCI patients and optimize AD treatments. We will make this clear in Conclusion.
Unclear Methodological Contribution (R1) Our main contribution is precisely defining ‘neurodegeneration dynamics’ with deep learning, rather than introducing novel DL methodology. We also established a method to derive subject-specific dynamics from public datasets and demonstrated effects of AD-related covariates on the dynamics. We validated our approach both quantitatively (Table 1) and qualitatively (Table 2). We anticipate this work will ignite neurodegeneration dynamics research.
Low Prediction Performance (R1) We found a minor error in RMSE calculation for the DISCLOSE family in Table 1. The corrected RMSEs of DISCLOSE are L1 only = (0.016, 0.039, 0.067, 0.089, 0.108), L1 + diff = (0.016, 0.040, 0.068, 0.090, 0.108), L1 + diff + Lasso = (0.018, 0.045, 0.078, 0.104, 0.127) at (τ < 1, 1, 2, 3, >3). The amended result is similar as before except vitrually identical short-term (τ < 1) prediction with SINDy. Adding dynamics difference or Lasso were not effective as before. This is not a new experiment and did not change both ablation study (+diff and +diff + Lasso were ineffective), conclusion (DISCLOSE was better than SINDy), and other data (Table 2). We apologize for the misunderstanding from our mistake.
Improper Loss Function Design (R2) The dynamics difference term in eq. (2) came from INSITE (Kacpryzk et al. 2024), and Lasso was for explicit sparsity. The INSITE paper also showed that adding the dynamics difference term was only effective in nonlinear ODE (Table 10, Kacpryzk et al. 2024) but not in a linear example. (2) could be effective in general.
Ineffective time windows (R3) Up to 15% of amnestic MCI patients progress to AD dementia over a 5-10 year period (Liss et al. 2021). We agree with R3 that accurate short-term prediction is important, but we want to also emphasize that long-term prediction is equally valuable for clinical planning and intervention assessment.
Miscellaneous (R1 – R3) (R1) We will disclose our in-house MRI processing pipeline and cite previous studies compared its performance with widely-used software (e.g. FreeSurfer) in the camera-ready version. (R2) Regarding missing INSITE performance, we attempted to evaluate it by applying SINDy twice (globally then individually). However, individually fitting with SINDy produced unstable parameters, so INSITE was omitted in Table 1. We will mention this in Section 4. (R3) For suggestion about functional connectivity, we agree with its importance and interest. Using rs-fMRIs in ADNI, finding the relationship between neurodegeneration dynamics and functional connectivity changes during AD continuum would be a promising next step. (R1, R2) We will clarify all acronyms, standardize numerical precision in tables, correct citation errors, use consistent variable notation, and number all equations. Again, we sincerely appreciate the reviewers’ focused proofreading and dedicated reviews to our work. These amendments would greatly improve the paper quality.
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’
The paper offers an interesting methodological contribution, and while some concerns were raised about clarity, experimental context, and the design of the loss function, the rebuttal addressed these points reasonably. After rebuttal to improve clarity and provide more context for the broader MICCAI audience, the work is suitable for publication. I recommend acceptance.