List of Papers Browse by Subject Areas Author List
Abstract
Accurate localization of myocardial infarction is essential for risk stratification. While LGE-MRI remains the gold standard, it is resource-intensive. Integrating cine MRI with ECG enables a more detailed representation of infarct properties. Existing inverse MI inference methods overlook realistic scar morphology and cardiac repolarization, reducing sensitivity to subtle ECG variations and interpretability of infarct-induced electrophysiological changes. In this paper, we propose a novel framework for noninvasive MI localization using cardiac digital twins. To bridge the domain gap between simulation and reality, we introduce an anatomy-aware stochastic infarct synthesis strategy to synthesize realistic, irregular scars with border zones, mimicking ischemic transmural progression. We then construct a virtual cohort to simulate QRS-T waveforms, capturing both depolarization and repolarization of electrophysiological dynamics. Furthermore, we design a Physiology and Anatomy Aware Network (PAA-Net) that jointly encodes 3D myocardial geometry and multi-lead ECGs to infer infarct area with various localizations, sizes, spatial extents and transmuralities. Experimental results demonstrate that our framework significantly outperforms existing methods in inverse inference, achieving 0.7391 and 0.5503 in Dice_Scar and Dice_BZ, respectively, and further enhances interpretability of the ECG–infarct relationship. The code is available at github.com/WangMengxiao319/PAA-Net-Cardiac-Digital-Twin.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/0066_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/WangMengxiao319/PAA-Net-Cardiac-Digital-Twin
Link to the Dataset(s)
N/A
BibTex
@InProceedings{WanMen_Physiology_MICCAI2026,
author = { Wang, Mengxiao AND Lyu, Yilin AND Camps, Julia AND Sia, Ching-Hui AND Chan, Mark Y. AND Jin, Yanrui AND Ge, Shuzhi Sam AND Liu, Chengliang AND Li, Lei},
title = { { Physiology and Anatomy Aware Inverse Inference of Myocardial Infarction for Cardiac Digital Twin } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16889},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The main contribution is the myocardial infarct localisation from a 12 lead ECG and a biventricular cardiac geometry leveraging synthetic infarcts generation and synthetic data simulation, in order to train a deep learning model.
- 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 different steps in the design of the model and the data processing take into account physiological knowledge on ECG data the digital twin construction and simulation is well designed.
- 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.
this method requires a 3D biventricular geometry built from MR imaging, but no motion, texture, or late enhancement information is used from images, therefore it would be hard to motivate MRI acquisition, the application is unclear. It would be interesting to know the performance drop when a generic heart is given. The external validation gives weak results. A Dice score of 0.22 is probably not very relevant.
- 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
Previous work on realistic synthetic infarct generation and digital twins for MI localisation could be cited. For instance this paper also designed a process for realistic infarct shape and digital twin simulation:
Nicolas Duchateau, Mathieu de Craene, Pascal Allain, Eric Saloux, and Maxime Sermesant. Infarct localization from myocardial deformation: Prediction and uncertainty quantification by regression from a low-dimensional space. IEEE Transactions on Medical Imaging, 35(10):2340-2352, 2016.
- 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 interesting topic and approach, but given the external validation score, the generalisation of the proposed method is unclear, and the fact that the method requires 3D imaging for cardiac shape but does not leverage any other image information is limiting clinical applicability. Performance in a more general setting should be evaluated.
- 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 author proposed an AI model for identifying myocardial infarction from ECG data and cardiac CINE MR imaging. The network consisted of an ECG encoder, combining a lead encoder and a signal encoder, and a biventricular geometry encoder based on the PointNet++ architecture. The ECG features were processed using a BiLSTM model and integrated into the geometry branch via feature-wise linear modulation. In addition, the coordinates of each ventricular point were represented using a four-dimensional Cobiveco biventricular coordinate system and used as geometric input to the PointNet++ model. The model was trained using a composite loss based on the predicted and ground-truth labels for three classes: scar, border zone, and healthy tissue. Training used cardiac simulation data generated from 130 meshes, producing 2210 simulated samples across 17 chronic infarction scenarios. Seventeen real ECG recordings were also included for external validation.
Overall, the proposed network showed good performance in identifying scar and border zone regions in simulated data. However, the differences between infarct categories remained substantial, and the model appeared to perform less well on the real external validation dataset.
- 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.
-
Good model architecture: the author proposes an architecture that effectively integrates information from the two modalities, ECG and MRI, in several scenarios. The proposed method also showed significant improvement over the baseline. The failure to identify the lateral small and inferior infarction regions, for example, may be attributed to the limited separation between the generated ECG signals. Although the evaluation on real data shows the lack of generalisability, this might be more closer realated to the training data simuation rather than the model itself.
-
The proposed method shows a significant improvement over the baseline approach.
-
- 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.
-
There is a lack of state-of-the-art comparison. The current work is only compared with another multimodal approach (cine CMR + ECG) for MI localisation, where both model not generalise well to real data. However, the monomodality approach referenced in citation 25 is reported to be more accurate and generalisable. Without a comparison with the method, which required the same or less data, the clinical relevance of the proposed multimodal approach remains limited.
-
The absence of standard deviations in the results table makes it difficult to assess the variability in accuracy across infarct categories. The lack of detail on the infarct subcategories in the external data also makes the results difficult to interpret.
-
The description of scar region simulation in Section 2.1 lacks sufficient methodological detail for reproducibility, particularly regarding λ, the range of τ base, and how Gaussian filtering of uniform noise generates a spatially correlated noise field.
-
The exclusion of right ventricular scar is not sufficiently justified despite its clinical relevance, especially in inferior infarction. It is unclear why RVMI was not or could not be included in the current pipeline. Its potential impact on biventricular EP simulation and forwarded ECG signals should be addressed.
-
- 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?
The methodology proposed in this work shows strong potential and may be of considerable interest for addressing inverse problems in the cardiac modelling community. However, the author should provide additional implementation details and a more thorough discussion of the current pipeline’s limitations.
- 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
First, they introduce an anatomy-aware stochastic infarct synthesis strategy that generates realistic infarct geometries with irregular shapes and border zones, better reflecting ischemic progression and improving the physiological fidelity of simulated data. Second, they construct a large virtual cohort incorporating electrophysiological heterogeneity, enabling the simulation of both depolarization and repolarization dynamics in QRS-T waveforms.
Building on this, they propose a novel Physiology and Anatomy Aware Network (PAA-Net) that jointly learns from 3D myocardial structure and multi-lead ECG signals to infer infarct location, extent, and transmurality. Unlike prior approaches, the proposed framework explicitly models the relationship between structural abnormalities and electrophysiological signatures, leading to improved interpretability.
- 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.
Novel physiology- and anatomy-informed formulation
Realistic infarct synthesis bridging simulation-to-reality gap
Incorporation of full electrophysiological dynamics (QRS–T)
Multimodal deep learning architecture (PAA-Net)
Improved interpretability of ECG–infarct relationship
Comprehensive and meaningful evaluation
Clear potential for clinical impact and scalability
- 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.
-Limited validation on real clinical data Although the paper is motivated by clinical applicability, most of the evaluation is conducted on simulated data generated from the proposed pipeline. The external validation set contains only 17 real cases, which is quite limited for establishing robustness and clinical utility. -Simulation-to-reality gap not fully demonstrated While the method aims to improve realism of infarct modeling, there is no direct validation showing that the synthetic scars match real clinical distributions. -Limited comparison to other methods Evaluation is mainly against a single baseline, making it difficult to fully assess improvements relative to the broader field. -Incomplete ablation of key contributions The impact of important components such as infarct synthesis and QRS–T simulation is not fully isolated and quantified
- 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.
- 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 paper presents a well-structured and technically solid framework for myocardial infarction localization by integrating cardiac digital twins, electrophysiological simulation, and multimodal learning. The combination of anatomy-aware infarct synthesis, QRS–T waveform simulation, and the proposed PAA-Net architecture is coherent and shows clear performance improvements over the baseline. The work is relevant to the MICCAI community and addresses an important problem in noninvasive cardiac assessment.
However, several factors limit the strength of the contribution. The validation is largely based on simulated data, with only limited real-world evaluation, making it difficult to assess clinical applicability. The novelty is also somewhat incremental, as the main contributions arise from integrating existing ideas rather than introducing a fundamentally new formulation. Additionally, the evaluation could be stronger with broader baselines and more detailed analysis of key components such as the simulation pipeline.
Overall, the paper is technically sound and shows promising results, but the level of validation and novelty is not strong enough for a clear accept. The final decision would depend on how well the authors address concerns regarding realism, generalization, and component-wise justification in the rebuttal.
- 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
We sincerely thank the reviewers for their valuable comments, constructive suggestions, and recognition of our work. Below, we’ll address key reviewer comments and will incorporate necessary revisions in the camera-ready version.
- Details in Method and Results. (R2-2; R2-3) We agree that these details are currently insufficiently described. In the camera-ready version, we will supplement the results table with standard deviations and provide clearer information about the infarct subcategories in the external dataset. We will also expand Section 2.1 with additional methodological details on scar region simulation, including the definition of λ, the range of τ_base, and the generation of spatially correlated noise fields using Gaussian-filtered uniform noise, to improve clarity and reproducibility.
- Incomplete ablation of key contributions (R3-4) The infarct synthesis process was designed based on the distribution patterns observed in real clinical data, and we have additionally conducted experiments comparing different synthesis strategies and their impact on performance, which will be included in the camera-ready version. In addition, the QRS–T simulation follows established prior methodologies, and within our pipeline we further quantified its effects through sensitivity analysis (Fig.2), demonstrating consistency with known clinical characteristics.
- Study design (R2-4). In our current study, the majority of the available and reliable clinical data primarily involved LV infarction; therefore, we focused on LV scar modeling as an initial step. We agree that RV involvement may affect biventricular activation and ECG signals, particularly in inferior infarction cases, and this limitation will be clarified. Future work will investigate incorporating RV scar characterization into the simulation pipeline.
- Simulation-to-reality gap (R1-2; R3-1; R3-2) We agree that the current external validation is limited and that a simulation-to-reality gap still exists. The primary goal of this work is to provide an initial physiology-aware infarct simulation framework rather than claim full clinical deployment. We believe the gap mainly arises from limited subject-specific EP personalization and temporal mismatch between MRI and ECG acquisitions in the external dataset. Nevertheless, the infarct synthesis process was designed based on real clinical distributions, and additional analyses comparing different synthesis strategies will be included in the camera-ready version. Future work will focus on more personalized EP simulations, temporally aligned ECG acquisition, and larger-scale clinical validation.
- Comparison with related work. (R2-1; R3-3) The method in Ref. 25 is based on 2D cine CMR slices with additional motion information extracted from 25-frame cine MRI sequences, and only considers short-axis slices, which differs from our single-frame setting and 3D mesh-based framework constructed from both long- and short-axis views. In addition, due to the specificity of this task, there are currently very limited directly comparable baselines. Nevertheless, we acknowledge the importance of broader comparisons and will further clarify these differences in the camera-ready version.
- Clinical application (R1-1). LGE MRI was used to determine scar location during external validation, while CINE MRI provided patient-specific 3D biventricular anatomy for EP simulation and inverse inference. We agree that the current framework does not yet fully utilize the rich MRI information, but it establishes a foundation for future incorporation of dynamic information and 4D mesh-based modeling.
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.
This paper presents a comprehensive framework for noninvasive myocardial infarction localisation using cardiac digital twins, with three main contributions: an anatomy-aware stochastic infarct synthesis strategy producing realistic irregular scars with border zones, a virtual cohort simulation capturing full QRS-T electrophysiological dynamics including repolarisation, and a novel PAA-Net architecture that jointly encodes 3D myocardial geometry and multi-lead ECG signals via a PointNet++ backbone conditioned on ECG features through FiLM modulation.
The approach is technically sound and well-presented with interesting methodology. However, the evaluation is dominated by simulated data, with very limited validation on real clinical cases, making it difficult to assess robustness and translational relevance. Method comparisons are narrow. Several aspects of the simulation pipeline and parameter choices are insufficiently detailed to ensure reproducibility, and the individual impact of major components is not fully isolated through ablation.
Despite these limitations, the reviewers unanimously agree that the papers novelty makes it worthy of acceptance, and I share the view that this work merits acceptance.