List of Papers Browse by Subject Areas Author List
Abstract
Given the global prevalence and high mortality of coronary artery disease (CAD), automated CAD diagnosis should evolve toward personalized methods to maximize its clinical value. However, existing techniques have been limited to artery-level prediction, lacking patient-level causality and failing to effectively account for individual patient confounders. In this work, for the first time, we introduce a Causal-Holistic Adaptive Intervention Network (CAIN) that tailors personalized CAD diagnosis for individual patients. CAIN generates semantic representations at both the patient and artery dual-levels for each case, constructing a holistic causal graph that captures individual-specific characteristics. It then implements adaptive causal intervention based on the patient’s specific condition, using dynamically updated and differentiated intervention variables. Experimental results on CCTA scans from 602 patients and 6,830 coronary branches across three clinical centers show that CAIN outperforms state-of-the-art methods, offering personalized clinical guidance. The source code is available at (***).
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/0169_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{MaXin_ACausalholistic_MICCAI2025,
author = { Ma, Xinghua AND Qiu, Xingyu AND Chu, Yuetan AND Wang, Kuanquan AND Qiu, Zhaowen AND Luo, Gongning AND Gao, Xin},
title = { { A Causal-holistic Adaptive Intervention Network for Tailoring Automated Coronary Artery Disease Diagnosis to Individual Patients } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15967},
month = {September},
page = {3 -- 13}
}
Reviews
Review #1
- Please describe the contribution of the paper
This study proposes a Causal-Holistic Adaptive Intervention Network designed to simultaneously perform patient-level diagnosis of CAD-RADS grade and artery-level stenosis localization. The key contribution lies in integrating global features extracted from entire CTA scans with localized information from individual arteries in CPR volumes, thereby enhancing the network’s diagnostic performance.
- 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 motivation to combine high-level features from whole coronary artery tree and low-level features from individual arteries is of novelty;
- Splitting data by clinical centers is a good way to evaluate models’ generability.
- 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 speed in the inference stage needs to be considered given that both CTA and CRP columns are fed into the model for feature extraction, along with a dynamic confounder bank.
- To extract features on the patient’s level, the whole CTA was partitioned into 3d cubics. It has some potential issues: a) 1. coronary arteries occupy a small portion of the whole cta scans, posing challenges to extracting useful/relevant features. b) the continuity and structure of the vessels can be damaged during the partitions.
- The ablation study is not comprehensive: a) to validate the effectiveness of patient-level configurations by removing Fcpr, Cat and patient-level prediction together is not appropriate. It remains unknown which factor plays the role in the performance drop. The model may work just fine with only artery-level features and patient-level prediction with no need for patient-level features.
- The model’s performances on different levels of disease severity are not reported. Authors can consider reporting the baseline characteristics of the testing dataset to give a hint.
- Sensitivity should be included in evaluation metrics.
- The loss equations need clearer descriptions. what does the subscript j represent? Symbols of features from different modules should be marked on the Figure, in line with the context.
- The authors stated in the abstract that a holistic causal graph is constructed. However, there is no graph structure in the CAIN.
- The study did not investigate deeply the so-called “causality or confounders” in the model predictions. What are the confounders hindering artery level-only predictions? What types of patient features is beneficial to the predictions?
- 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 has provided an anonymized link to the source code, dataset, or any other dependencies.
- 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?
This study is novel in its model design, but no comprehensive ablation study was conducted to explain the relationship between the various designs and answer questions about confounding factors and causality.
- 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
Review #2
- Please describe the contribution of the paper
The paper presents a new model named CAIN, which integrates patient-level and artery-level information into a causal graph for coronary artery disease diagnosis. Compared to traditional approaches, CAIN incorporates a causal graph that includes voxel-level features derived from curved planar reformation, confounders, and a causal inference algorithm for diagnosis. Experimental results indicate consistent improvements with CAIN in coronary artery stenosis detection, plaque characterization, and CAD-RADS classification.
- 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 major strength of the paper is its use of a holistic causal graph to link voxel-level features with artery-level features, thereby enabling patient-level diagnostic prediction.
- 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 major weakness of the paper lies in
- The sensitivity to the size of the confounder bank (M) or the hyperparameter beta beyond just one test.
- It’s unclear whether the gains come from causal modeling or just from using patient-level information in any form.
- The confounder bank and its update mechanism (Eq. 3) assume that cosine similarity effectively identifies relevant confounding variables. However, this may not capture the full complexity of confounding in CCTA, especially across different scanners or patient anatomies. If confounders are inaccurately modeled, the causal inference may become unreliable, leading to suboptimal or biased predictions.
- While causal framing is used, it’s not clear how clinical practitioners can trace the rationale behind specific predictions, particularly at the patient level.
- What does the function ‘do’ in Eq.2 indicate?
- 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.
(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 holistic causal graph is novel, but the resulting predictions are difficult to interpret, making it challenging to convince clinical practitioners. In addition, all hyperparameters—including M, beta, etc.—were set to fixed values, and no discussion or analysis was provided regarding their impact
- 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.
The authors addressed all my concerns.
Review #3
- Please describe the contribution of the paper
The authors propose a novel, holistic approach for computer-aided diagnosis (CAD) of coronary artery disease (CAD) from Coronary CT Angiography (CCTA) images, based on principles of causal AI. Utilizing multi-scale representations and a dynamically updated causal intervention scheme, they construct a structural causal model (SCM) that captures artery- and patient-level characteristics free from the effects of imaging confounders. These multi-scale embeddings are integrated through a latent space association framework, and confounders are isolated via causal interventions that leverage a dynamically updated database of confounded representations. To the best of my knowledge, this is the first study to apply causal AI to address the challenge of dual-level prediction—integrating artery-based and patient-level information. While prior work has only recently begun incorporating patient-level features into artery-based prediction models, this study advances that integration. The authors validate their approach on a private, retrospective dataset of 602 scans from multiple clinical centers, demonstrating clear performance improvements over existing methods that rely solely on artery-based predictions combined with conventional patient-level models.
- 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.
Methodology: The methodology is highly innovative. To the best of my knowledge, this is the first study to combine multi-scale information from coronary branch arteries and patient characteristics within a comprehensive causal AI framework. The method is successfully developed to provide patient-level diagnosis based on the CAD-RADS clinical standard. The method also provides explanations for different types of lesions at the vessel level. This approach can be extended to other imaging modalities and applications, promoting the advancement of personalized medicine.
Validation: The method was validated downstream on the task of patient level diagnosis and artery based lesion detection based on a retrospective dataset from different clinical centers, with each center being used as a different split (center 1/2 for training/validation, center 3 for the testing) independent from each others, demonstrating the generalization capabilities of the approach. Quantitatively, the method clearly outperforms existing methods in the literature across six different metrics (ACC, Precision/Recall, F1, NPV, Specificity), with the performance differences being statistically significant (p-value) on both tasks.
- 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.
Clarity: Given the limited space the authors tried to include all the necessary information to explain the causal AI aspects of the methodology. The work would benefit on further expansion on aspects and details such as the dynamically updated intervention method. Methodology: The authors consider a limited size for the confounder bank storing maximum 1000 vectors, and they dedicate a considerable section of their method on a way to aggregate the vectors under limited storage space constraints. This is not a bottleneck in the method as this is a very small amount that can be stored in memory. A recent technology that can be used called vector databases can store millions of vectors effortlessly.
The authors do not discuss the limitations of the method in cases of pathology at the patient as well as artery level and the future steps
- 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
1) In the section 2.2 more details are needed to explain the composition of the artery-prediction head. What this head involves what is its architecture? Lot of notations (m, e) are not visible in Figure 2.
2) Minor corrections: On page 2 shouldn’t the arrow go from confounder (d) to imaging singal of branch (x)? If it is correct in the next then probably its direction is wrong in figure 1. In figure 1 the arrow directions are barely visible. Also, in the dataset description in section 3 it is not LXC instead of LCX branch? Also on Table 1 please correct No-stensis to Stenosis.
3) It is not clear if the artery-level feature extractor Fcpr extracts embeddings from all available arteries or parts, and if part what is the combination.
4) Why the maximum capacity M of B was set to 1000? To me it seems a very low number there are vector databases (AWS open search, Weaviate) that can store and query millions of vectors effortlessly.
5) What is the training and inference time of the method and how it fits in the clinical context of CAD
- 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 paper introduces a highly innovative and well-executed methodology grounded in causal AI, which is the first, for coronary artery disease diagnosis, a clinically significant and technically challenging problem. The integration of multi-scale artery and patient-level representations into a structural causal model is not only novel but also demonstrates impressive generalization across clinical centers. Despite minor areas where additional clarity would enhance the presentation, the contribution is substantial. I strongly recommend acceptance and nominate this work for a best paper. Also it would benefit tremendously from expanding to a journal to let the authors further analyze their approach.
- 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.
The paper demonstrates significant methodological contributions that would be worth of presenting in the MICCAI conference. After the rebuttal my opinion has not change about the quality of the work.
Author Feedback
We sincerely appreciate your positive feedback on our work, including its novelty (“is novel”-R1; “highly innovative”-R2; “is of novelty”-R3) and validation (“consistent improvements”-R1; “clear performance improvements”-R2; “a good way to evaluate”-R3). We also thank you for taking the time to read our paper and provide constructive suggestions. We will address your concerns as follows:
R1:
1) Regarding hyperparameter sensitivity, CAIN performs stably as M ranges from 1000 to 1500 and beta ranges from 0.8 to 0.95, with the best performance at beta=0.9, which is therefore selected. 2) CAIN’s performance (Fig.4c) drops notably when the scale of causal modeling is reduced, despite the presence of patient-level information, clearly confirming the effectiveness of causal modeling. 3) *Cosine similarity effectively identifies confounding variables. This has been validated by comparing it with other measures (e.g., Manhattan distance and Pearson correlation). *Confounder modeling addresses complexity across scanners and patient anatomies, achieved through dynamic updates during case learning. 4) The recent clinical study (Faro, JACC, 2023) has demonstrated associations among coronary branches. CAIN’s patient-level design enables rational prediction based on these findings, supporting clinical credibility. 5) The term ‘Do’ refers to the do-expression (Wang, CVPR, 2020), which denotes the intervention operation.
R2:
*We will incorporate your suggestions to provide more detailed results and analyses in the code-link, and further optimize our method in the extended study. *Although CAIN offers clear advantages, it still relies on centerline extraction, which is similar to that used in previous methods, making it susceptible to errors during the extraction process. *We plan to explore vector database technology with larger confounder banks and design a more robust mechanism to eliminate the above dependency.
R3:
1) Although both CTA and CPRs require feature extraction, the total number of extractions needed to predict all branches for a patient is comparable to that of SOTAs and does not affect inference speed. The confounder bank is constructed during training and thus does not impact inference efficiency. 2) The shift-window mechanism used in patient-level representation effectively models the continuity and structure of the vessels across patches, while the hierarchical attention mechanism enhances relevant feature extraction. 3) *F_cpr, C_at, and patient-level prediction are removed together, as they collectively constitute the patient-level pipeline. *The necessity of patient-level features is evidenced by a significant performance drop (Fig.4c) as the number of branches involved in the intervention approaches one. 4) Tab.1 reports the mean and variance across different levels of disease severity to reflect overall performance. A more detailed report will be available via the code-link. 5) Sensitivity is computed in the same way as Recall (Tab.1). 6) *j in L_loc and L_char denotes the index of e_qry and the lesion category, respectively, in each loss term. *The symbols for features will be further annotated in the final version. 7) The causal graph (Fig.1b) aligns with CAIN’s architecture: artery-level representation and prediction correspond to X->Y_a, while patient-level prediction, based on patient-level representation and artery-level prediction, correspond to X->Y_p and Y_a->Y_p. The causal intervention removes spurious associations related to D. 8) *As noted on Page 2, Line 27~31, clinical studies [2,11] indicate that confounders originate from both external factors (e.g., data heterogeneity) and internal factors (e.g., lesion-independent information). Center-based validation and qualitative results demonstrate CAIN’s effectiveness in handling both types. *Features most relevant to the current patient are beneficial to the predictions and are adaptively matched through a dynamic update strategy and causal intervention.
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’
N/A