List of Papers Browse by Subject Areas Author List
Abstract
Interpretable models are crucial for supporting clinical decision-making, driving advances in their development and application for
medical images. However, the nature of 3D volumetric data makes it inherently challenging to visualize and interpret intricate and complex structures like the cerebral cortex. Cortical surface renderings, on the other hand, provide a more accessible and understandable 3D representation of brain anatomy, facilitating visualization and interactive exploration. Motivated by this advantage and the widespread use of surface data for studying neurological disorders, we present the eXplainable Surface Vision Transformer (X-SiT). This is the first inherently interpretable neural network that offers human-understandable predictions based on interpretable cortical features. As part of X-SiT, we introduce a prototypical surface patch decoder for classifying surface patch embeddings, incorporating case-based reasoning with spatially corresponding cortical prototypes. The results demonstrate state-of-the-art performance in detecting Alzheimer’s disease and frontotemporal dementia while additionally providing informative prototypes that align with known disease patterns and reveal classification errors.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/0394_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)
ADNI dataset: https://adni.loni.usc.edu
NIFD dataset: https://memory.ucsf.edu/research-trials/research/allftd
BibTex
@InProceedings{BonFab_XSiT_MICCAI2025,
author = { Bongratz, Fabian and Wolf, Tom Nuno and Gual Ramon, Jaume and Wachinger, Christian},
title = { { X-SiT: Inherently Interpretable Surface Vision Transformers for Dementia Diagnosis } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15973},
month = {September},
page = {692 -- 702}
}
Reviews
Review #1
- Please describe the contribution of the paper
They provide a prototype-based method to explain the results of cortical surface transformer. By applying to ADNI and NIFD, they verify they achieved more interpretable and quantitatively comparable results compared to the previous methods.
- 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 explainable is a very popular and important topic in medical image computing area, while it is hardly explored for cortical surface based models in the literature.
- They combined the prototype-based method in computer vision with some prior knowledge of cortical surface, i.e., the cortical surfaces are typically well aligned in most scaeniro.
- They applied the method for disease diagnosis, which is clinically important.
- The paper is well organized and pretty clear.
- 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 feature choice in experiment is kind of uncommon. The sulcal depth and curvature are folding-based features, which are typically used for tasks, such as surface registration and parcellation. Meanwhile, to diagnoize the cognition-related diseases, such as AD, I supposed it is more common to use some behaviour-related cortical features like, thickness as they used in this work, and surface area and volume.
-
Some citations are incorrect. In experiment settings, it should cite reference no. 35 for spherical u-net, not for reference no. 4.
-
More discussion is expected. As they mentioned in experimental results, only 76.3% and 71.7% critical regions are overlapped for 5 randomly initialization models. Is it satisfactory? What if they used an ensemble strategy to used the overlapped regions to guide their adaptive surface patch scaling weights, would it achieve more robust and more accurate results? It would be interesting if they could investigate in the reproducibility of the interpretable results using multi-institutional datasets.
-
What are the advantages of the prototype-based methods over the other interpretable methods, such as saliency map-based methods, should be explained.
-
It is a minor point the diagnosis accuracy decreases a little bit after applying the interpretable method.
-
- 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?
I think the research focus of this work is important, their method is reasonable, results are satisfactory overall, and the organization is clear. I agreed that this work is above the boarderline for acceptance.
- 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 primary contribution of this paper is the development of X-SiT, an inherently interpretable vision transformer for cortical surface analysis. X-SiT integrates interpretability directly into its architecture through the Prototypical Surface Patch (PSP) Decoder. This allows it to classify brain surfaces based on learned, spatially-aligned prototypes that correspond to known disease patterns.
- 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.
Although prototype learning based on feature similarity is not new, its application to the surface data modality is novel.
- 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 performance of the proposed model is not optimal among baselines. Also, it would be better if the paper had included MCI subjects in the experiments.
- 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 paper proposes an interpretable surface transformer model for disease analysis. The method is novel and shows promising results.
- 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 #3
- Please describe the contribution of the paper
This paper proposes a novel surface-based Transformer framework for cortical disease classification, incorporating a prototype-based decoder to enhance model interpretability. The model operates directly on the native mesh without requiring spherical mapping or complex resampling. A sparse scaling strategy enables efficient training while emphasizing task-relevant regions. The method achieves state-of-the-art performance on ADNI and NIFD datasets (AD vs. FTD classification), and the learned prototype activations highlight spatially meaningful regions based on the similarity between input features and representative prototypes.
- 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 works directly on the native surface without requiring spherical mapping, improving practicality. The prototype-based decoder is structurally interpretable and produces clearly visualizable activation maps. The sparse scaling mechanism enables efficient learning by focusing on informative regions. The method achieves competitive results compared to SOTA methods, while providing greater explainability.
- 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 framework lacks an ablation study to evaluate the individual contributions of the sparse scaling and prototype decoder components. The interpretability analysis is limited to a few visual examples (e.g., two figures) and lacks systematic or quantitative evaluation.
- 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
The prototype-based interpretability strategy has the potential to be extended to group-level analyses or to incorporate clinical variables (e.g., biomarker scores) as input features. This would be a promising direction for future work.
- 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?
This paper proposes a prototype-based decoder design that explicitly incorporates interpretability into a surface-based cortical classification framework. While many explainable models suffer from reduced accuracy, this method achieves strong interpretability while maintaining classification performance comparable to prior approaches. The absence of ablation studies and the use of a limited set of disease types and datasets indicate areas for improvement in future work. Nonetheless, the careful architectural design, balanced experimental setup, and strong results justify an overall recommendation for acceptance.
- 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