pith. sign in

arxiv: 2605.22044 · v1 · pith:JSRZY2GWnew · submitted 2026-05-21 · 💻 cs.CV

Physiology and Anatomy Aware Inverse Inference of Myocardial Infarction for Cardiac Digital Twin

Mengxiao Wang , Yilin Lyu , Julia Camps , Ching Hui Sia , Mark Yan-Yee Chan , Yanrui Jin , Shuzhi Sam Ge , Chengliang Liu
show 1 more author
Lei Li
This is my paper

Pith reviewed 2026-05-22 06:36 UTC · model grok-4.3

classification 💻 cs.CV
keywords myocardial infarctioncardiac digital twininverse inferencescar segmentationECGcine MRIphysiology-aware modelingborder zone
0
0 comments X

The pith

A digital twin approach infers myocardial infarction locations from cine MRI and ECG by synthesizing realistic scars and jointly processing heart geometry with electrical signals.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a noninvasive method for locating heart muscle damage by building cardiac digital twins that combine cine MRI geometry with ECG readings. It creates a virtual patient cohort using an anatomy-aware strategy to generate irregular scars with border zones that follow realistic ischemic patterns, then simulates the resulting QRS-T waveforms to capture both contraction and recovery phases. A Physiology and Anatomy Aware Network is trained on these pairs to map input geometry and multi-lead signals directly to infarct regions of different sizes and depths. This setup closes the gap between simulated training data and real patient recordings, yielding higher accuracy than prior inverse methods while making the link between ECG changes and scar properties more traceable.

Core claim

The central claim is that an anatomy-aware stochastic infarct synthesis method can produce realistic scars and border zones whose simulated QRS-T waveforms, when paired with 3D myocardial geometry, train a Physiology and Anatomy Aware Network to segment scar and border-zone regions from combined cine-MRI and multi-lead ECG inputs at Dice scores of 0.7391 and 0.5503, outperforming existing inverse inference techniques and increasing interpretability of infarct-induced electrophysiological changes.

What carries the argument

The Physiology and Anatomy Aware Network (PAA-Net) jointly encodes 3D myocardial geometry from cine MRI and multi-lead ECG signals, trained on a virtual cohort generated by anatomy-aware stochastic infarct synthesis that creates irregular transmural scars with border zones and corresponding QRS-T waveforms.

If this is right

  • Noninvasive MI localization becomes feasible using only widely available cine MRI and ECG rather than resource-intensive LGE-MRI.
  • Sensitivity to subtle ECG variations increases because training data includes realistic scar morphology and repolarization dynamics.
  • The relationship between specific infarct properties and observed ECG changes gains clearer mechanistic mapping.
  • Risk stratification for post-infarction patients can incorporate detailed scar extent and transmurality estimates derived from routine recordings.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Extending the same synthesis and joint-encoding approach to other ECG-altering conditions such as hypertrophy or bundle-branch block could produce a unified cardiac digital-twin pipeline.
  • Real-time fusion of the inferred infarct map with wearable ECG streams might allow continuous monitoring of evolving scar-related arrhythmia risk.
  • If the simulation-reality gap is further reduced by adding fiber-orientation or perfusion data, the same framework could support patient-specific therapy planning.

Load-bearing premise

The anatomy-aware stochastic infarct synthesis strategy produces scars that accurately mimic real ischemic transmural progression and the simulated QRS-T waveforms capture real depolarization and repolarization dynamics sufficiently to bridge the simulation-reality domain gap.

What would settle it

A blinded comparison of the network's predicted scar and border-zone maps against contemporaneous LGE-MRI ground truth in an independent clinical cohort of at least 50 patients with varying infarct locations would directly test whether the reported Dice improvements hold on real data.

Figures

Figures reproduced from arXiv: 2605.22044 by Chengliang Liu, Ching Hui Sia, Julia Camps, Lei Li, Mark Yan-Yee Chan, Mengxiao Wang, Shuzhi Sam Ge, Yanrui Jin, Yilin Lyu.

Figure 1
Figure 1. Figure 1: The pipeline of cohort simulation and inverse inference of infarcted area. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sensitive analysis for synthetic MI scenarios. (a) Mutual dissimilarity [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: External validation visu￾alization [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: ECG saliency maps and qualitative infarct localization visualization. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original 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 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 areas with varying localizations, sizes, spatial extents, and transmuralities. Experimental results demonstrate that our framework significantly outperforms existing methods in inverse inference, achieving Dice scores of 0.7391 and 0.5503 for scar and border zone segmentation, respectively, while further enhancing the interpretability of the ECG-infarct relationship. Our code will be released upon acceptance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes a framework for noninvasive myocardial infarction (MI) localization via cardiac digital twins. It introduces an anatomy-aware stochastic infarct synthesis strategy to generate realistic irregular scars with border zones mimicking ischemic transmural progression, constructs a virtual cohort to simulate QRS-T waveforms capturing depolarization and repolarization dynamics, and designs a Physiology and Anatomy Aware Network (PAA-Net) that jointly encodes 3D myocardial geometry and multi-lead ECGs to infer infarct areas of varying localizations, sizes, extents, and transmuralities. The central claim is that this approach significantly outperforms existing inverse inference methods, achieving Dice scores of 0.7391 for scar and 0.5503 for border zone segmentation while improving interpretability of the ECG-infarct relationship.

Significance. If the simulation-reality gap is successfully bridged and the reported performance generalizes, the work would be significant for enabling detailed, non-invasive MI characterization that integrates cine MRI and ECG, potentially reducing dependence on resource-intensive LGE-MRI for risk stratification. The joint geometry-ECG encoding and stochastic synthesis for irregular scars with border zones address important limitations in prior inverse methods. Releasing code upon acceptance would support reproducibility in the cardiac digital twin community.

major comments (2)
  1. [Methods section describing the anatomy-aware stochastic infarct synthesis strategy] Methods section describing the anatomy-aware stochastic infarct synthesis strategy: the claim that this produces scars whose geometry, border-zone extent, and transmural progression statistically match real ischemic scars is load-bearing for the domain-gap-bridging assertion and for the Dice scores being predictive of real inverse-inference performance, yet no quantitative shape or distribution comparisons (e.g., scar volume histograms, border-zone width statistics, or transmurality profiles) against a held-out real LGE-MRI cohort are supplied.
  2. [Experimental results section reporting Dice scores of 0.7391 and 0.5503] Experimental results section reporting Dice scores of 0.7391 and 0.5503: the outperformance claim requires that the virtual-cohort QRS-T simulations reproduce real depolarization/repolarization effects seen in multi-lead clinical ECGs, but no waveform distribution tests (e.g., QRS duration, T-wave morphology statistics) or direct validation against paired real ECG/LGE-MRI data are provided, leaving open the possibility that PAA-Net overfits to simulation artifacts.
minor comments (1)
  1. [Abstract] The abstract states that the framework 'further enhanc[es] the interpretability of the ECG-infarct relationship' without indicating the quantitative or qualitative metric used to demonstrate this improvement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive review of our manuscript. We address the major comments point by point in the following responses. We believe these revisions will strengthen the paper.

read point-by-point responses

  1. Referee: [Methods section describing the anatomy-aware stochastic infarct synthesis strategy] Methods section describing the anatomy-aware stochastic infarct synthesis strategy: the claim that this produces scars whose geometry, border-zone extent, and transmural progression statistically match real ischemic scars is load-bearing for the domain-gap-bridging assertion and for the Dice scores being predictive of real inverse-inference performance, yet no quantitative shape or distribution comparisons (e.g., scar volume histograms, border-zone width statistics, or transmurality profiles) against a held-out real LGE-MRI cohort are supplied.

    Authors: We thank the referee for this insightful comment. The anatomy-aware stochastic infarct synthesis strategy is designed to generate irregular scar geometries with border zones by incorporating anatomical information from the 3D myocardial geometry and modeling the stochastic nature of ischemic progression. This approach is motivated by physiological knowledge rather than claiming exact statistical equivalence to real scars. We agree that providing quantitative comparisons would be beneficial. In the revised manuscript, we will add supplementary figures and tables showing distributions of scar volumes, border zone widths, and transmurality profiles from our synthetic scars, along with comparisons to statistics reported in the clinical literature on LGE-MRI. This will help substantiate the realism of the synthesized data. revision: partial

  2. Referee: [Experimental results section reporting Dice scores of 0.7391 and 0.5503] Experimental results section reporting Dice scores of 0.7391 and 0.5503: the outperformance claim requires that the virtual-cohort QRS-T simulations reproduce real depolarization/repolarization effects seen in multi-lead clinical ECGs, but no waveform distribution tests (e.g., QRS duration, T-wave morphology statistics) or direct validation against paired real ECG/LGE-MRI data are provided, leaving open the possibility that PAA-Net overfits to simulation artifacts.

    Authors: We appreciate the referee's concern about the fidelity of the simulated ECGs and potential overfitting. Our QRS-T simulations utilize a detailed forward model based on the bidomain equations with realistic fiber orientations and action potential dynamics to capture both depolarization and repolarization phases. The virtual cohort includes a wide range of infarct configurations to ensure diversity. To mitigate concerns of simulation-specific artifacts, we will include in the revision additional analyses comparing ECG waveform statistics, such as QRS duration and T-wave morphology metrics, from our simulated data to those observed in clinical ECG repositories. While direct validation with paired real ECG and LGE-MRI data is not currently presented due to the focus on digital twins and limited access to such comprehensive paired datasets, the superior performance on held-out simulated cases with varied parameters supports the generalizability within this framework. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent simulation assumptions

full rationale

The paper presents a framework that generates synthetic infarct data via an anatomy-aware stochastic synthesis strategy and simulates QRS-T waveforms on virtual cohorts to train the PAA-Net for inverse inference. No equations, fitted parameters renamed as predictions, or self-citation chains are exhibited in the text that reduce the core claims (Dice scores or ECG-infarct interpretability) to the inputs by construction. The synthesis and simulation steps are described as external data-generation processes whose validity is independent of the network's learned mappings, rendering the overall derivation self-contained.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim relies on the effectiveness of the proposed synthesis strategy and network architecture, with training parameters fitted to simulated data.

free parameters (1)
  • neural network weights in PAA-Net
    Fitted during training on simulated virtual cohort data to map geometry and ECG to infarct predictions.
axioms (1)
  • domain assumption Simulated data from digital twins can approximate real cardiac electrophysiology and anatomy for training purposes.
    Invoked in the construction of virtual cohort and training of the network to bridge simulation-reality gap.
invented entities (1)
  • PAA-Net no independent evidence
    purpose: Joint encoding of 3D myocardial geometry and multi-lead ECGs for infarct inference.
    New network architecture proposed to improve sensitivity to subtle ECG variations.

pith-pipeline@v0.9.0 · 5782 in / 1344 out tokens · 63976 ms · 2026-05-22T06:36:44.289653+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

  1. [1]

    Medical Image Analysis p

    Camps, J., Wang, Z.J., Doste, R., Berg, L.A., Holmes, M., Lawson, B., Tomek, J., Burrage, K., Bueno-Orovio, A., Rodriguez, B.: Harnessing 12-lead ECG and MRI data to personalise repolarisation profiles in cardiac digital twin models for enhanced virtual drug testing. Medical Image Analysis p. 103361 (2024)

    work page 2024

  2. [2]

    Annals of Internal Medicine114(4), 264–270 (1991)

    Christian, T.F., Clements, I.P., Behrenbeck, T., Huber, K.C., Chesebro, J.H., Gersh, B.J., Gibbons, R.J.: Limitations of the Electrocardiogram in Estimating Infarction Size after Acute Reperfusion Therapy for Myocardial Infarction. Annals of Internal Medicine114(4), 264–270 (1991)

    work page 1991

  3. [3]

    European Heart Journal41(48), 4556–4564 (2020)

    Corral-Acero, J., Margara, F., Marciniak, M., Rodero, C., Loncaric, F., Feng, Y., Gilbert, A., Fernandes, J.F., Bukhari, H.A., Wajdan, A., Martinez, M.V., San- tos, M.S., Shamohammdi, M., Luo, H., Westphal, P., Leeson, P., DiAchille, P., Gurev, V., Mayr, M., Geris, L., Pathmanathan, P., Morrison, T., Cornelussen, R., Prinzen, F., Delhaas, T., Doltra, A., ...

    work page 2020

  4. [4]

    In: Func- tional Imaging and Modeling of the Heart

    Dillon, J.R., Mauger, C., Zhao, D., Deng, Y., Petersen, S.E., McCulloch, A.D., Young, A.A., Nash, M.P.: An Open-Source End-to-End Pipeline for Generating 3D+t Biventricular Meshes from Cardiac Magnetic Resonance Imaging. In: Func- tional Imaging and Modeling of the Heart. pp. 372–383. Springer Nature Switzer- land, Cham (2025)

    work page 2025

  5. [5]

    Doste, R., Camps, J., Wang, Z.J., Berg, L.A., Holmes, M., Smith, H., Beetz, M., Li, L., Banerjee, A., Grau, V., Rodriguez, B.: An Automated Computational Pipeline for Generating Large-Scale Cohorts of Patient-Specific Ventricular Models in Elec- tromechanical In Silico Trials (2025)

    work page 2025

  6. [6]

    IEEE Reviews in Biomedical Engineering18, 316–336 (2025) 10 Wang et al

    Li, L., Camps, J., Rodriguez, B., Grau, V.: Solving the Inverse Problem of Elec- trocardiography for Cardiac Digital Twins: A Survey. IEEE Reviews in Biomedical Engineering18, 316–336 (2025) 10 Wang et al

    work page 2025

  7. [7]

    IEEE Transactions on Medical Imaging pp

    Li, L., Camps, J., Wang, Z., Beetz, M., Banerjee, A., Rodriguez, B., Grau, V.: To- wards Enabling Cardiac Digital Twins of Myocardial Infarction Using Deep Com- putational Models for Inverse Inference. IEEE Transactions on Medical Imaging pp. 1–1 (2024)

    work page 2024

  8. [8]

    Advanced Science 12(30), e06933 (2025)

    Mehdi, R.R., Kadivar, N., Mukherjee, T., Mendiola, E.A., Bersali, A., Shah, D.J., Karniadakis, G., Avazmohammadi, R.: Non-Invasive Diagnosis of Chronic Myocar- dial Infarction via Composite In-Silico-Human Data Learning. Advanced Science 12(30), e06933 (2025)

    work page 2025

  9. [9]

    Interna- tional Journal of Epidemiology40(1), 139–146 (2011)

    Mendis, S., Thygesen, K., Kuulasmaa, K., Giampaoli, S., Mähönen, M., Ngu Black- ett, K., Lisheng, L., Writing group on behalf of the participating experts of the WHO consultation for revision of WHO definition of myocardial infarction: World Health Organization definition of myocardial infarction: 2008–09 revision. Interna- tional Journal of Epidemiology4...

    work page 2008

  10. [10]

    In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2021

    Mu, L., Liu, H.: Cardiac Transmembrane Potential Imaging with GCN Based Iterative Soft Threshold Network. In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2021. pp. 547–556. Springer International Pub- lishing, Cham (2021)

    work page 2021

  11. [11]

    Journal of Computational Physics346, 191–211 (2017)

    Neic, A., Campos, F.O., Prassl, A.J., Niederer, S.A., Bishop, M.J., Vigmond, E.J., Plank, G.: Efficient computation of electrograms and ECGs in human whole heart simulations using a reaction-eikonal model. Journal of Computational Physics346, 191–211 (2017)

    work page 2017

  12. [12]

    Nature Reviews Cardiology16(2), 100–111 (2019)

    Niederer, S.A., Lumens, J., Trayanova, N.A.: Computational models in cardiology. Nature Reviews Cardiology16(2), 100–111 (2019)

    work page 2019

  13. [13]

    Current Cardiology Reviews10(3), 229–236 (2014)

    Nikus, K., Birnbaum, Y., Eskola, M., Sclarovsky, S., Zhong-qun, Z., Pahlm, O.: Up- dated Electrocardiographic Classification of Acute Coronary Syndromes. Current Cardiology Reviews10(3), 229–236 (2014)

    work page 2014

  14. [14]

    In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2025

    Peng, J., Beetz, M., Banerjee, A., Chen, M., Grau, V.: An Anatomical Significance- Aware Architecture for Explainable Myocardial Infarction Prediction via Multi- task Learning. In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2025. pp. 13–23. Springer Nature Switzerland, Cham (2026)

    work page 2025

  15. [15]

    Proceedings of the AAAI Conference on Artificial Intelligence32(1) (2018)

    Perez, E., Strub, F., De Vries, H., Dumoulin, V., Courville, A.: FiLM: Visual Rea- soning with a General Conditioning Layer. Proceedings of the AAAI Conference on Artificial Intelligence32(1) (2018)

    work page 2018

  16. [16]

    Qi, C.R., Yi, L., Su, H., Guibas, L.J.: PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space (2017)

    work page 2017

  17. [17]

    In: Artificial Intelligence in Healthcare

    Ramzan, F., Kiberu, Y., Jathanna, N., Jamil-Copley, S., Clayton, R.H., Chen, C.: CLAIM: Clinically-Guided LGE Augmentation for Realistic and Diverse Myocar- dial Scar Synthesis and Segmentation. In: Artificial Intelligence in Healthcare. pp. 279–292. Springer Nature Switzerland, Cham (2026)

    work page 2026

  18. [18]

    Medical Image Analysis74, 102247 (2021)

    Schuler, S., Pilia, N., Potyagaylo, D., Loewe, A.: Cobiveco: Consistent biventricular coordinates for precise and intuitive description of position in the heart – with MATLAB implementation. Medical Image Analysis74, 102247 (2021)

    work page 2021

  19. [19]

    Neurophotonics7(1), 015008 (2020)

    Tran, A.P., Yan, S., Fang, Q.: Improving model-based functional near-infrared spectroscopy analysis using mesh-based anatomical and light-transport models. Neurophotonics7(1), 015008 (2020)

    work page 2020

  20. [20]

    Circulation: Cardiovas- cular Imaging p

    Trankle, C.R., Jordan, J.H.: Synthetic Contrast-Free LGE in Acute MI: Assessing the Promise and Boundaries of Diffusion-Based Modeling. Circulation: Cardiovas- cular Imaging p. e019473 (2026)

    work page 2026

  21. [21]

    Information Fusion123, 103302 (2025) Physiology and Anatomy Aware Inverse MI Inference 11

    Wang, M., Li, Z., Tian, Y., Wei, X., Jin, Y., Liu, C.: BioCross: A cross-modal framework for unified representation of multi-modal biosignals with heterogeneous metadata fusion. Information Fusion123, 103302 (2025) Physiology and Anatomy Aware Inverse MI Inference 11

    work page 2025

  22. [22]

    2024.Global status report on alcohol and health and treatment of substance use disorders

    World Health Organization: World health statistics 2021: Mon- itoring health for the SDGs, sustainable development goals. https://www.who.int/publications/i/item/9789240027053 (2021)

    work page arXiv 2021

  23. [23]

    Medical Image Analysis62, 101668 (2020)

    Xu, C., Xu, L., Ohorodnyk, P., Roth, M., Chen, B., Li, S.: Contrast agent-free synthesis and segmentation of ischemic heart disease images using progressive se- quential causal GANs. Medical Image Analysis62, 101668 (2020)

    work page 2020

  24. [24]

    In: Medical Image Comput- ing and Computer-Assisted Intervention – MICCAI 2014, vol

    Xu, J., Sapp, J.L., Rahimi Dehaghani, A., Gao, F., Wang, L.: Variational Bayesian Electrophysiological Imaging of Myocardial Infarction. In: Medical Image Comput- ing and Computer-Assisted Intervention – MICCAI 2014, vol. 8674, pp. 529–537. Springer International Publishing, Cham (2014)

    work page 2014

  25. [25]

    Radiology291(3), 606–617 (2019)

    Zhang, N., Yang, G., Gao, Z., Xu, C., Zhang, Y., Shi, R., Keegan, J., Xu, L., Zhang, H., Fan, Z., Firmin, D.: Deep Learning for Diagnosis of Chronic Myocardial Infarction on Nonenhanced Cardiac Cine MRI. Radiology291(3), 606–617 (2019)

    work page 2019

  26. [26]

    eLife13, RP93002 (2024)

    Zhou, X., Wang, Z.J., Camps, J., Tomek, J., Santiago, A., Quintanas, A., Vazquez, M., Vaseghi, M., Rodriguez, B.: Clinical phenotypes in acute and chronic infarction explained through human ventricular electromechanical modelling and simulations. eLife13, RP93002 (2024)

    work page 2024