List of Papers Browse by Subject Areas Author List
Abstract
Automatic disentanglement of progression subtypes in neurodegenerative disorders is critical for patient stratification in clinical management and trial enrichment. Existing disease subtyping methods operate exclusively on low-dimensional scalar biomarkers, discarding the fine-grained spatial patterns necessary for biologically meaningful subtype discovery and patient-level personalization. Conversely, continuous-time models that directly operate on longitudinal 3D brain MRI recover a single shared population trajectory and perform no subtyping, critically limiting personalized prediction and clinical interpretability. To fill this gap, we present STAMP, a longitudinal progression subtyping framework that, integrates automatic subtype discovery directly into the optimization of a continuous-time spatio-temporal atlas from longitudinal 3D T1 brain MRI. STAMP jointly learns subtype-specific temporal placements, dense image deformation fields, and region-wise progression functions, enabling probabilistic subtype assignment and personalized trajectory prediction for new subjects. Evaluated on longitudinal ADNI-1/GO/2 data, STAMP achieves the lowest forecasting error in 16 of 18 anatomical regions and statistically significant improvements in 11 regions compared to single-trajectory and post-hoc subtyping baselines, while recovering clinically interpretable subtype structure.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/3598_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://gitlab.inria.fr/jkalkhof/stamp
Link to the Dataset(s)
https://adni.loni.usc.edu/data-samples/adni-data/
BibTex
@InProceedings{KalJoh_STAMP_MICCAI2026,
author = { Kalkhof, John AND Lorenzi, Marco},
title = { { STAMP: Subtype-Aware Spatio-Temporal Atlases for Personalized Prediction of Longitudinal Alzheimer’s Disease Progression } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16895},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
This paper proposes a longitudinal progression subtyping framework named STAMP that integrates automatic subtype discovery into the optimization of a continuous-time spatio-temporal atlas from longitudinal 3D brain MRI for personalized trajectory prediction.
- 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 overall research topic is highly valuable because longitudinal analysis and the generation of disease trajectories represent crucial frontiers in the medical imaging community. 2 The method reasonably incorporates neurological priors to guide regional progression analysis, resulting in a clinically meaningful contribution that should be extended to broader disease categories and morphological topics in future studies. 3 The writing is clear, and the figure is well presented.
- 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 While the modeling is logical, real-world brain disease evolution is highly non-linear, and the sequential process of first analyzing subtype progression to subsequently drive the deformation backbone might introduce overly strong linear assumptions and error accumulation, which the authors need to clarify again. 2 The authors are encouraged to explore advanced medical world model architectures like Brain-WM [1] to utilize a unified multimodal framework that can collaboratively process subtype discovery and spatio-temporal deformation, allowing the two tasks to mutually align and promote each other end-to-end. 3The study lacks validation on independent external longitudinal cohorts, making it difficult to confirm whether the discovered subtypes are generalizable neurodegenerative patterns or merely overfitted to the specific data distribution of the ADNI dataset.
[1] Brain-WM: Brain Glioblastoma World Model, arxiv, 2026.
- 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.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/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 research topic is interesting. The methodology is relatively novel. The experimental results are relatively sufficient.
- 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
The paper proposes a subtype-aware longitudinal disease progression model that jointly learns the subtype structure and continuous-time spatio-temporal atlas directly using 3D T1 brain MRI scans. The method can infer the subtype membership and personalized future trajectory simultaneously for given subjects. Experimental results on longitudinal ADNI-1/GO/2 data showcase lower predicting errors in majority of regions and also statistically significant improvements compared to single-trajectory baselines, while providing with interpretable subtypes.
- 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 overcomes the limitations of the existing subtyping work which utilize scalar biomarkers such as the SuStaIn framework, and also improves upon imaging-based disease progression modeling without tackling the heterogeneity, therefore this paper is well-positioned to bridge the gap and potentially contribute to the field of neurogenerative disease modeling and subtyping
-
The key contribution of the paper is the novel integration of subtyping framework with continuous-time atlas modeling, which is conceptually strong as the subtype assignment is embedded in an end-to-end fashion within the model by a joint learning scheme
-
Besides the combination of several technical components such as Neural ODE-based progression functions, the authors further apply regularization to avoid all subjects collapse into trivial solutions, as this is a critical issue for subtyping in this domain
-
The experiments are relatively comprehensive with the comparison against multiple imaging-based progression models, and the ablation study is also well-designed to demonstrate the contribution of each component of STAMP such as subtype modeling
-
- 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 evaluation metrics are limited as the subtype clustering quality is not directly evaluated, for example, there are metrics such as ARI or NMI could be applied for evaluating the subtyping quality
-
The current experiment lacks the comparison to subtyping methods in the domain, as the paper is somewhat positioned against methods like SuStaIn
-
There is limited discussion on cases where the subtype assignment fails or the uncertainty in the subtyping, as well as the sensitivity to the number of subtypes
-
All the experiments are only conducted on ADNI dataset, without external validation dataset such as BLSA, AIBL, OASIS, etc., which limits the generalizability and robustness 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/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 — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This is an overall well-written paper with novel contributions to the field of neurodegenerative disease modeling and subtyping, by bridging the gap between these two topics. This could potentially inspire future work for disease modeling and parsing disease heterogeneity.
- 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 paper introduces a spatiotemporal disease progression model, designed for brain MRI and neurodegenerative disease, that accounts for distinct disease subtypes. This fills a clear hole in current disease progression modelling technology, where current subtype models do not consider the whole image or continuous temporal change and current spatiotemporal models do not consider subtypes. Results are encouraging showing improved prediction of longitudinal morphological change over a well-chosen set of contemporary baselines.
- 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.
-
Fills a clear gap in disease progression modelling technology
-
The implementation is well grounded in recent modelling advances.
-
The paper is clear and well explained.
-
The results are sufficient to show efficacy of the new idea
-
- 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.
-
These preliminary results don’t really show disease subtypes but rather seem to separate out people on a trajectory of normal aging (subtype 1) from those on a genuine dementia trajectory (subtype 2). This is evidenced by the fact that most of the CN cases get assigned to subtype 1 while all of the dementia cases get assigned to subtype 2. This does have some value, but does not quite achieve what the authors intend the method to achieve. Perhaps some further tuning of the parameters might reveal actual disease subtypes, as found using e.g. SuStaIn on the same data set.
-
Figure 2 isn’t particularly useful as presented. It is hard to distinguish the subtypes by eye. More useful would be to show a visual comparison with SuStaIn subtypes to highlight whether the STAMP subtypes align at corresponding disease/model stages
-
- 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.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- 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/2026/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.
(6) Strong Accept — must be accepted due to excellence
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This is excellent work - the strongest in my pile. Makes a clear advance in disease progression modelling - an area of high activity and potential impact. While the results remain preliminary, the method is well constructed and clearly on the right lines for the next phase of development in this area.
- 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
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
Reviewers summarized strengths of your work, novel aspects and noted that your solution may be of potential interest. Major strengths include the novel incorporation of neurological priors into the longitudinal modeling of disease progression, its clinically meaningful contribution, resulting interpretable subtypes and comprehensive comparison, and the clear writing. Reported weaknesses address the issue of incorporating the expected highly non-linear disease progression, the lack of validation on external longitudinal cohorts (an important issue since the paper focuses on ADNI only), and somewhat limited evaluation metrics. Would the paper be accepted for MICCAI, authors are strongly encouraged to provide a revision that integrates suggestions for improvements that consider specific weaknesses as discussed in the reviews. In particular Rev#3 provides an important discussion on “disease subtypes” versus differences in aging trajectories.