pith. sign in

arxiv: 2502.11378 · v1 · submitted 2025-02-17 · 📡 eess.SP

Numerical Differentiation-based Electrophysiology-Aware Adaptive ResNet for Inverse ECG Modeling

Lingzhen Zhu , Kenneth Bilchick , Jianxin Xie This is my paper

Pith reviewed 2026-05-23 03:21 UTC · model grok-4.3

classification 📡 eess.SP
keywords inverse ECGelectrocardiographic imagingelectrophysiology constraintsresidual neural networknumerical differentiationadaptive networkcardiac mapping
0
0 comments X

The pith

Numerical differentiation over local regions lets an adaptive residual network enforce electrophysiology constraints more effectively for inverse ECG reconstruction.

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

The paper develops a neural network method to reconstruct electrical activity on the heart surface from body-surface ECG measurements. It applies numerical differentiation so that electrophysiology laws are enforced across small local patches in space and time instead of globally. An adaptive residual network structure is added to maintain stable gradient flow during training and reduce sensitivity to starting weights. These changes target known failure modes in earlier physics-informed models such as overfitting and poor scaling. If the approach holds, heart-surface potential maps become more accurate and consistent even when body measurements contain typical noise levels.

Core claim

Numerical differentiation is used to obtain spatiotemporal derivatives that allow electrophysiology constraints to act over local regions rather than the entire domain, while an adaptive residual network improves gradient propagation and initialization; together these changes produce more accurate and stable solutions to the inverse ECG problem than prior methods.

What carries the argument

EAND-ARN, which computes local spatiotemporal derivatives via numerical differentiation to strengthen EP constraint application and employs adaptive residual blocks to improve gradient flow during training.

If this is right

  • Reconstructed cardiac electrical maps exhibit lower error under realistic measurement noise.
  • The network avoids overfitting to global EP constraints while remaining scalable to deeper architectures.
  • Training converges reliably from varied initial weights without manual tuning.

Where Pith is reading between the lines

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

  • The local-constraint technique could transfer to other inverse problems that combine data with partial differential equation models.
  • Clinical workflows might incorporate the method for real-time arrhythmia localization if inference speed remains comparable to existing tools.
  • Further tests on multi-patient cohorts with varying torso geometries would reveal whether the local enforcement generalizes beyond the training distribution.

Load-bearing premise

Numerical differentiation applied over local spatiotemporal regions strengthens EP constraint enforcement without introducing approximation errors that degrade reconstruction accuracy or cause overfitting.

What would settle it

A controlled test on synthetic data with known ground-truth heart potentials where the new method is run both with and without the local numerical differentiation step; if reconstruction error does not decrease when the local step is present, the central claim is falsified.

Figures

Figures reproduced from arXiv: 2502.11378 by Jianxin Xie, Kenneth Bilchick, Lingzhen Zhu.

Figure 1
Figure 1. Figure 1: Illustration of the proposed EAND-ARN framework for cardiac electrodynamic prediction. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of Adap-ResBlock. adaptive residual block is shown in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) 2D square grid; (b) 3D triangle mesh on human heart surface [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The temporal derivative at ut derived from four temporal neighbors. ut−2τ = ut − 2τ ∂ut ∂t + (2τ ) 2 2! ∂ 2ut ∂t2 − (2τ ) 3 3! ∂ 3ut ∂t3 + (2τ ) 4 4! ∂ 4ut ∂t4 + O(τ 5 ) ut−τ = ut − τ ∂ut ∂t + τ 2 2! ∂ 2ut ∂t2 − τ 3 3! ∂ 3ut ∂t3 + τ 4 4! ∂ 4ut ∂t4 + O(τ 5 ) ut+τ = ut + τ ∂ut ∂t + τ 2 2! ∂ 2ut ∂t2 + τ 3 3! ∂ 3ut ∂t3 + τ 4 4! ∂ 4ut ∂t4 + O(τ 5 ) ut+2τ = ut + 2τ ∂ut ∂t + (2τ ) 2 2! ∂ 2ut ∂t2 + (2τ ) 3 3! ∂ 3u… view at source ↗
Figure 5
Figure 5. Figure 5: Experimental design for model performance evaluation. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (a) RE, (b) MSE, and (c) CC yielded by EP-DL, EP-DL∆ND , and EP-DLND. (a) (b) (c) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The influence of number of neurons on (a) [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The influence of different neural network structures on (a) RE, (b) MSE, and (c) CC. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The comparison of methods (i.e., Tikh 2nd, STRE, EP-DL, and EAND-ARN) on (a) RE, (b) MSE, and (c) CC under different BSP noise levels (σ = 0.01, 0.05, and 0.1) electrical potential pattern from the anterior view of the heart. The Tikh 2nd and STRE approaches remain to exibit non￾negligible differences compared with the reference in [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Spatial visualization of reference HSP u, and the predicted HSP uˆ yielded by methods including Tikh 2nd, STRE, EP-DL, and EAND-ARN with different observation noise level σ = 0.01, 0.05, 0.1; (a) the basal view; (b) the anterior view. (a) (b) (c) [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The comparison of estimated HSP evolvement over time at a specific spatial location with noise level (a) 0.01, (b) 0.05, and (c) 0.1 [PITH_FULL_IMAGE:figures/full_fig_p012_11.png] view at source ↗
read the original abstract

Electrocardiographic imaging aims to noninvasively reconstruct the electrical dynamic patterns on the heart surface from body-surface ECG measurements, aiding the mechanistic study of cardiac function. At the core of ECGI lies the inverse ECG problem, a mathematically ill-conditioned challenge where small body measurement errors or noise can lead to significant inaccuracies in the reconstructed heart-surface potentials. %Leveraging a well-developed electrophysiological (EP) model, our previous study developed an EP-informed deep learning framework, demonstrating promising effectiveness in improving cardiac map predictions. To improve the accuracy of ECGI and ensure that cardiac predictions adhere to established physical principles, recent advances have incorporated well-established electrophysiology (EP) laws into their model formulations. However, traditional EP-informed models encounter significant challenges, including overfitting to EP constraints, limitations in network scalability, and suboptimal initialization. These issues compromise prediction accuracy and stability, hindering their effectiveness in practical applications. This highlights the need for an advanced data analytic and predictive tool to achieve reliable cardiac electrodynamic restoration. Here, we present a Numerical Differentiation-based Electrophysiology-Aware Adaptive Residual neural Network (EAND-ARN) for robust inverse ECG modeling. Our method employs numerical differentiation to compute the spatiotemporal derivative, enabling EP constraints to be applied across a local spatiotemporal region, thereby strengthening the overall EP enforcement. Additionally, we design an adaptive residual network to improve gradient flow, enhancing predictive accuracy and mitigating issues with poor initialization. Experimental results show that EAND-ARN significantly outperforms existing methods in current practice.

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 / 2 minor

Summary. The manuscript proposes EAND-ARN, a Numerical Differentiation-based Electrophysiology-Aware Adaptive Residual neural Network for the inverse ECG problem. It claims that applying EP constraints via numerical differentiation over local spatiotemporal regions strengthens enforcement, while an adaptive residual network improves gradient flow and mitigates poor initialization. The central claim is that experimental results demonstrate significant outperformance over existing methods in current practice.

Significance. If the experimental outperformance holds under rigorous validation, the approach could advance ECGI by providing a scalable way to embed EP physics without the overfitting or initialization issues of prior EP-informed models. The adaptive residual design and local spatiotemporal constraint application are potentially useful extensions of prior work, but significance is limited by the absence of any reported handling of noise amplification in the differentiation step.

major comments (2)
  1. [Methods (numerical differentiation)] Methods section (description of numerical differentiation): the claim that applying EP constraints 'across a local spatiotemporal region' strengthens enforcement without introducing approximation errors is load-bearing for the central claim, yet no differentiation order, stencil size, or noise-robust variant (e.g., Savitzky-Golay) is specified. Finite-difference differentiation is first-order accurate and noise-amplifying; without explicit mitigation, the reported outperformance cannot be attributed to stronger EP enforcement rather than overfitting to noisy derivatives.
  2. [Experimental results] Experimental results section: the abstract asserts that EAND-ARN 'significantly outperforms existing methods,' but the provided text supplies no datasets, cross-validation protocol, error bars, or ablation isolating the numerical differentiation component. This prevents assessment of whether the local-region EP constraint actually improves reconstruction accuracy or merely fits noise.
minor comments (2)
  1. [Abstract] Abstract: the sentence 'Leveraging a well-developed electrophysiological (EP) model, our previous study...' appears to be a remnant from an earlier draft and should be removed or integrated.
  2. [Abstract] Notation: 'EAND-ARN' is introduced without expanding the acronym on first use in the abstract.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the manuscript accordingly to improve methodological transparency and experimental rigor.

read point-by-point responses

  1. Referee: [Methods (numerical differentiation)] Methods section (description of numerical differentiation): the claim that applying EP constraints 'across a local spatiotemporal region' strengthens enforcement without introducing approximation errors is load-bearing for the central claim, yet no differentiation order, stencil size, or noise-robust variant (e.g., Savitzky-Golay) is specified. Finite-difference differentiation is first-order accurate and noise-amplifying; without explicit mitigation, the reported outperformance cannot be attributed to stronger EP enforcement rather than overfitting to noisy derivatives.

    Authors: We agree that the Methods section requires explicit specification of the numerical differentiation scheme to support the central claims. The current manuscript does not detail the differentiation order, stencil size, or any noise-robust variant. In the revised version, we will expand this section to specify these parameters (including the finite-difference order and stencil) and describe incorporation of a noise-mitigation approach such as Savitzky-Golay filtering. This will allow clearer attribution of performance gains to the local spatiotemporal EP enforcement. revision: yes

  2. Referee: [Experimental results] Experimental results section: the abstract asserts that EAND-ARN 'significantly outperforms existing methods,' but the provided text supplies no datasets, cross-validation protocol, error bars, or ablation isolating the numerical differentiation component. This prevents assessment of whether the local-region EP constraint actually improves reconstruction accuracy or merely fits noise.

    Authors: We concur that the Experimental results section lacks the necessary details for rigorous evaluation. The manuscript will be revised to include descriptions of the datasets, cross-validation protocol, error bars or statistical significance measures, and an ablation study isolating the numerical differentiation component. These additions will enable assessment of whether the local EP constraints contribute to improved accuracy beyond noise fitting. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper introduces EAND-ARN as a new architecture combining numerical differentiation for local EP constraint enforcement with an adaptive residual network. Claims rest on experimental outperformance rather than any derivation that reduces by construction to inputs, fitted parameters renamed as predictions, or self-citation chains. The abstract references prior EP-informed work but treats it as background, not as a load-bearing uniqueness theorem or ansatz that forces the current result. No equations or steps in the provided text exhibit self-definitional reduction or renaming of known results as novel derivations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no free parameters, invented entities, or explicit axioms are stated. The central modeling choice is the use of numerical differentiation to localize EP constraints.

axioms (1)
  • domain assumption Numerical differentiation accurately approximates the spatiotemporal derivatives needed to apply EP laws over local regions.
    Invoked to enable stronger constraint enforcement without point-wise application.

pith-pipeline@v0.9.0 · 5800 in / 988 out tokens · 43838 ms · 2026-05-23T03:21:32.756802+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

54 extracted references · 54 canonical work pages

  1. [1]

    Spatiotemporal differentiation of myocardial infarctions,

    H. Yang, C. Kan, G. Liu, and Y . Chen, “Spatiotemporal differentiation of myocardial infarctions,” IEEE Transactions on Automation Science and Engineering, vol. 10, no. 4, pp. 938–947, 2013

    work page 2013

  2. [2]

    Optimal electrocardiographic leads for detecting acute myocardial ischemia,

    B. M. Horacek, J. W. Warren, C. J. Penney, R. S. MacLeod, L. Title, M. Gardner, and C. Feldman, “Optimal electrocardiographic leads for detecting acute myocardial ischemia,” Journal of Electrocardiology , vol. 34, pp. 97–112, 2001

    work page 2001

  3. [3]

    An xml format for storing body surface potential map recordings,

    R. Bond, D. Finlay, C. Nugent, and G. Moore, “An xml format for storing body surface potential map recordings,” in 2009 36th Annual Computers in Cardiology Conference (CinC) . IEEE, 2009, pp. 233– 236

    work page 2009

  4. [4]

    Evaluation of arrhythmic causes of syncope: correlation between holter monitoring, electrophysiologic testing, and body surface potential map- ping,

    D. Lacroix, M. Dubuc, T. Kus, P. Savard, M. Shenasa, and R. Nadeau, “Evaluation of arrhythmic causes of syncope: correlation between holter monitoring, electrophysiologic testing, and body surface potential map- ping,” American heart journal , vol. 122, no. 5, pp. 1346–1354, 1991

    work page 1991

  5. [5]

    Noninvasive electrocardiographic imaging,

    Y . Rudy and J. E. Burnes, “Noninvasive electrocardiographic imaging,” Annals of Noninvasive electrocardiology , vol. 4, no. 3, pp. 340–359, 1999

    work page 1999

  6. [6]

    Optimal sensor placement for space–time potential mapping and data fusion,

    R. Zhu, B. Yao, H. Yang, and F. Leonelli, “Optimal sensor placement for space–time potential mapping and data fusion,” IEEE sensors letters, vol. 3, no. 1, pp. 1–4, 2018

    work page 2018

  7. [7]

    Sparse modeling and recursive prediction of space–time dynamics in stochastic sensor networks,

    Y . Chen and H. Yang, “Sparse modeling and recursive prediction of space–time dynamics in stochastic sensor networks,” IEEE Transactions on Automation Science and Engineering , vol. 13, no. 1, pp. 215–226, 2015

    work page 2015

  8. [8]

    Noninvasive electrocardiographic imaging: reconstruction of epicardial potentials, electrograms, and isochrones and localization of single and multiple electrocardiac events,

    H. S. Oster, B. Taccardi, R. L. Lux, P. R. Ershler, and Y . Rudy, “Noninvasive electrocardiographic imaging: reconstruction of epicardial potentials, electrograms, and isochrones and localization of single and multiple electrocardiac events,” Circulation, vol. 96, no. 3, pp. 1012– 1024, 1997

    work page 1997

  9. [9]

    The forward and inverse problems of electrocardiog- raphy,

    R. M. Gulrajani, “The forward and inverse problems of electrocardiog- raphy,” IEEE Engineering in Medicine and Biology Magazine , vol. 17, no. 5, pp. 84–101, 1998

    work page 1998

  10. [10]

    Mesh resolution impacts the accuracy of inverse and forward ecg problems,

    B. Yao, S. Pei, and H. Yang, “Mesh resolution impacts the accuracy of inverse and forward ecg problems,” in 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE, 2016, pp. 4047–4050

    work page 2016

  11. [11]

    Spatiotemporal regularization for inverse ecg mod- eling,

    B. Yao and H. Yang, “Spatiotemporal regularization for inverse ecg mod- eling,” IISE Transactions on Healthcare Systems Engineering , vol. 11, no. 1, pp. 11–23, 2020

    work page 2020

  12. [12]

    Recent progress in inverse prob- lems in electrocardiology,

    R. S. MacLeod and D. H. Brooks, “Recent progress in inverse prob- lems in electrocardiology,” IEEE Engineering in Medicine and Biology Magazine, vol. 17, no. 1, pp. 73–83, 1998

    work page 1998

  13. [13]

    Solution of physics-based bayesian inverse problems with deep generative priors,

    D. V . Patel, D. Ray, and A. A. Oberai, “Solution of physics-based bayesian inverse problems with deep generative priors,” Computer Methods in Applied Mechanics and Engineering , vol. 400, p. 115428, 2022

    work page 2022

  14. [14]

    Nonin- vasive electrocardiographic imaging for cardiac electrophysiology and arrhythmia,

    C. Ramanathan, R. N. Ghanem, P. Jia, K. Ryu, and Y . Rudy, “Nonin- vasive electrocardiographic imaging for cardiac electrophysiology and arrhythmia,” Nature medicine, vol. 10, no. 4, pp. 422–428, 2004

    work page 2004

  15. [15]

    Solutions of ill posed problems (v. h. winston and sons, distributed by wiley, new york),

    A. Tikhonov and V . Arsenin, “Solutions of ill posed problems (v. h. winston and sons, distributed by wiley, new york),” 1977

    work page 1977

  16. [16]

    Physics-driven spatiotemporal regularization for high-dimensional predictive modeling: A novel approach to solve the inverse ecg problem,

    B. Yao and H. Yang, “Physics-driven spatiotemporal regularization for high-dimensional predictive modeling: A novel approach to solve the inverse ecg problem,” Scientific reports, vol. 6, no. 1, p. 39012, 2016

    work page 2016

  17. [17]

    Inter-model consistency and complementarity: Learning from ex-vivo imaging and electrophysiological data towards an integrated understanding of cardiac physiology,

    O. Camara, M. Sermesant, P. Lamata, L. Wang, M. Pop, J. Relan, M. De Craene, H. Delingette, H. Liu, S. Niederer et al. , “Inter-model consistency and complementarity: Learning from ex-vivo imaging and electrophysiological data towards an integrated understanding of cardiac physiology,” Progress in Biophysics and Molecular Biology , vol. 107, no. 1, pp. 12...

    work page 2011

  18. [18]

    An electromechanical model of the heart for image analysis and simulation,

    M. Sermesant, H. Delingette, and N. Ayache, “An electromechanical model of the heart for image analysis and simulation,”IEEE transactions on medical imaging , vol. 25, no. 5, pp. 612–625, 2006

    work page 2006

  19. [19]

    Physics-constrained deep learning for robust inverse ecg modeling,

    J. Xie and B. Yao, “Physics-constrained deep learning for robust inverse ecg modeling,” IEEE Transactions on Automation Science and Engineering, vol. 20, no. 1, pp. 151–166, 2022

    work page 2022

  20. [20]

    Physics-informed neural networks for cardiac activation mapping,

    F. Sahli Costabal, Y . Yang, P. Perdikaris, D. E. Hurtado, and E. Kuhl, “Physics-informed neural networks for cardiac activation mapping,” Frontiers in Physics, vol. 8, p. 42, 2020

    work page 2020

  21. [21]

    Piratenets: Physics-informed deep learning with residual adaptive networks

    S. Wang, B. Li, Y . Chen, and P. Perdikaris, “Piratenets: Physics- informed deep learning with residual adaptive networks,” arXiv preprint arXiv:2402.00326, 2024

    work page arXiv 2024

  22. [22]

    Relating epicardial to body surface potential distributions by means of transfer coefficients based on geometry measurements,

    R. C. Barr, M. Ramsey, and M. S. Spach, “Relating epicardial to body surface potential distributions by means of transfer coefficients based on geometry measurements,” IEEE Transactions on biomedical engineering, no. 1, pp. 1–11, 1977

    work page 1977

  23. [23]

    An expert review of the inverse problem in electrocardiographic imaging for the non- invasive identification of atrial fibrillation drivers,

    Z. Yadan, L. Jian, W. Jian, L. Yifu, L. Haiying et al., “An expert review of the inverse problem in electrocardiographic imaging for the non- invasive identification of atrial fibrillation drivers,” Computer Methods and Programs in Biomedicine , vol. 240, p. 107676, 2023

    work page 2023

  24. [24]

    Resolution strategies for the finite-element-based solution of the ecg inverse problem,

    D. Wang, R. M. Kirby, and C. R. Johnson, “Resolution strategies for the finite-element-based solution of the ecg inverse problem,” IEEE Transactions on biomedical engineering , vol. 57, no. 2, pp. 220–237, 2009

    work page 2009

  25. [25]

    Ecg imaging to detect the site of ventricular ischemia using torso electrodes: a computational study,

    V . Kara, H. Ni, E. A. Perez Alday, and H. Zhang, “Ecg imaging to detect the site of ventricular ischemia using torso electrodes: a computational study,” Frontiers in physiology, vol. 10, p. 50, 2019

    work page 2019

  26. [26]

    The impact of torso signal processing on noninvasive electrocardiographic imaging reconstructions,

    L. R. Bear, Y . S. Dogrusoz, W. Good, J. Svehlikova, J. Coll-Font, E. Van Dam, and R. MacLeod, “The impact of torso signal processing on noninvasive electrocardiographic imaging reconstructions,” IEEE Transactions on Biomedical Engineering , vol. 68, no. 2, pp. 436–447, 2020

    work page 2020

  27. [27]

    Reducing line-of-block artifacts in cardiac activation maps estimated using ecg imaging: A comparison of source models and estimation methods,

    S. Schuler, M. Schaufelberger, L. R. Bear, J. A. Bergquist, M. J. Cluitmans, J. Coll-Font, ¨O. N. Onak, B. Zenger, A. Loewe, R. S. MacLeod et al. , “Reducing line-of-block artifacts in cardiac activation maps estimated using ecg imaging: A comparison of source models and estimation methods,” IEEE Transactions on Biomedical Engineering , vol. 69, no. 6, pp...

    work page 2041

  28. [28]

    The effect of interpolating low amplitude leads on the inverse reconstruction of cardiac electrical activity,

    A. S. Rababah, L. R. Bear, Y . S. Dogrusoz, W. Good, J. Bergquist, J. Stoks, R. MacLeod, K. Rjoob, M. Jennings, J. Mclaughlin et al., “The effect of interpolating low amplitude leads on the inverse reconstruction of cardiac electrical activity,” Computers in biology and medicine , vol. 136, p. 104666, 2021

    work page 2021

  29. [29]

    A new spatiotemporal regularization approach for reconstruction of cardiac transmembrane potential patterns,

    B. Messnarz, B. Tilg, R. Modre, G. Fischer, and F. Hanser, “A new spatiotemporal regularization approach for reconstruction of cardiac transmembrane potential patterns,” IEEE transactions on Biomedical Engineering, vol. 51, no. 2, pp. 273–281, 2004

    work page 2004

  30. [30]

    Physiological- model-constrained noninvasive reconstruction of volumetric myocardial 14 transmembrane potentials,

    L. Wang, H. Zhang, K. C. Wong, H. Liu, and P. Shi, “Physiological- model-constrained noninvasive reconstruction of volumetric myocardial 14 transmembrane potentials,” IEEE Transactions on Biomedical Engineer- ing, vol. 57, no. 2, pp. 296–315, 2009

    work page 2009

  31. [31]

    Hybrid neural state- space modeling for supervised and unsupervised electrocardiographic imaging,

    X. Jiang, R. Missel, M. Toloubidokhti, K. Gillette, A. J. Prassl, G. Plank, B. M. Hor ´aˇcek, J. L. Sapp, and L. Wang, “Hybrid neural state- space modeling for supervised and unsupervised electrocardiographic imaging,” IEEE Transactions on Medical Imaging , 2024

    work page 2024

  32. [32]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational physics , vol. 378, pp. 686–707, 2019

    work page 2019

  33. [33]

    Physics- informed neural networks (pinns) for fluid mechanics: A review,

    S. Cai, Z. Mao, Z. Wang, M. Yin, and G. E. Karniadakis, “Physics- informed neural networks (pinns) for fluid mechanics: A review,” Acta Mechanica Sinica, vol. 37, no. 12, pp. 1727–1738, 2021

    work page 2021

  34. [34]

    Physics-informed deep learning for com- putational elastodynamics without labeled data,

    C. Rao, H. Sun, and Y . Liu, “Physics-informed deep learning for com- putational elastodynamics without labeled data,” Journal of Engineering Mechanics, vol. 147, no. 8, p. 04021043, 2021

    work page 2021

  35. [35]

    Ep-pinns: Cardiac electrophysiology characterisation using physics-informed neural networks,

    C. Herrero Martin, A. Oved, R. A. Chowdhury, E. Ullmann, N. S. Peters, A. A. Bharath, and M. Varela, “Ep-pinns: Cardiac electrophysiology characterisation using physics-informed neural networks,” Frontiers in Cardiovascular Medicine, vol. 8, p. 768419, 2022

    work page 2022

  36. [36]

    Physics-informed neural networks for brain hemodynamic predictions using medical imaging,

    M. Sarabian, H. Babaee, and K. Laksari, “Physics-informed neural networks for brain hemodynamic predictions using medical imaging,” IEEE transactions on medical imaging , vol. 41, no. 9, pp. 2285–2303, 2022

    work page 2022

  37. [37]

    Physics-informed neural networks (pinns) for 4d hemodynamics prediction: an investigation of optimal framework based on vascular morphology,

    X. Zhang, B. Mao, Y . Che, J. Kang, M. Luo, A. Qiao, Y . Liu, H. Anzai, M. Ohta, Y . Guoet al., “Physics-informed neural networks (pinns) for 4d hemodynamics prediction: an investigation of optimal framework based on vascular morphology,”Computers in Biology and Medicine, vol. 164, p. 107287, 2023

    work page 2023

  38. [38]

    Cardiovascular care with digital twin technology in the era of generative artificial intelligence,

    P. M. Thangaraj, S. H. Benson, E. K. Oikonomou, F. W. Asselbergs, and R. Khera, “Cardiovascular care with digital twin technology in the era of generative artificial intelligence,” European Heart Journal , vol. 45, no. 45, pp. 4808–4821, 2024

    work page 2024

  39. [39]

    A spatial- temporally adaptive pinn framework for 3d bi-ventricular electrophysi- ological simulations and parameter inference,

    Y . Ye, H. Liu, X. Jiang, M. Toloubidokhti, and L. Wang, “A spatial- temporally adaptive pinn framework for 3d bi-ventricular electrophysi- ological simulations and parameter inference,” in International Confer- ence on Medical Image Computing and Computer-Assisted Intervention . Springer, 2023, pp. 163–172

    work page 2023

  40. [40]

    Automatic differentiation in machine learning: a survey,

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, “Automatic differentiation in machine learning: a survey,” Journal of machine learning research, vol. 18, no. 153, pp. 1–43, 2018

    work page 2018

  41. [41]

    Can-pinn: A fast physics-informed neural network based on coupled-automatic– numerical differentiation method,

    P.-H. Chiu, J. C. Wong, C. Ooi, M. H. Dao, and Y .-S. Ong, “Can-pinn: A fast physics-informed neural network based on coupled-automatic– numerical differentiation method,” Computer Methods in Applied Me- chanics and Engineering , vol. 395, p. 114909, 2022

    work page 2022

  42. [42]

    Dgm: A deep learning algorithm for solving partial differential equations,

    J. Sirignano and K. Spiliopoulos, “Dgm: A deep learning algorithm for solving partial differential equations,” Journal of computational physics, vol. 375, pp. 1339–1364, 2018

    work page 2018

  43. [43]

    Anderson, J

    D. Anderson, J. C. Tannehill, R. H. Pletcher, R. Munipalli, and V . Shankar, Computational fluid mechanics and heat transfer . CRC press, 2020

    work page 2020

  44. [44]

    Scientific machine learning through physics–informed neural networks: Where we are and what’s next,

    S. Cuomo, V . S. Di Cola, F. Giampaolo, G. Rozza, M. Raissi, and F. Piccialli, “Scientific machine learning through physics–informed neural networks: Where we are and what’s next,” Journal of Scientific Computing, vol. 92, no. 3, p. 88, 2022

    work page 2022

  45. [45]

    A unified deep artificial neural network approach to partial differential equations in complex geometries,

    J. Berg and K. Nystr ¨om, “A unified deep artificial neural network approach to partial differential equations in complex geometries,” Neu- rocomputing, vol. 317, pp. 28–41, 2018

    work page 2018

  46. [46]

    A simple two-variable model of cardiac excitation,

    R. R. Aliev and A. V . Panfilov, “A simple two-variable model of cardiac excitation,”Chaos, Solitons & Fractals, vol. 7, no. 3, pp. 293–301, 1996

    work page 1996

  47. [47]

    Review on classification based on artificial neural networks,

    K. Saravanan and S. Sasithra, “Review on classification based on artificial neural networks,” Int J Ambient Syst Appl , vol. 2, no. 4, pp. 11–18, 2014

    work page 2014

  48. [48]

    State-of-the-art in artificial neural network applications: A survey,

    O. I. Abiodun, A. Jantan, A. E. Omolara, K. V . Dada, N. A. Mohamed, and H. Arshad, “State-of-the-art in artificial neural network applications: A survey,” Heliyon, vol. 4, no. 11, 2018

    work page 2018

  49. [49]

    Deep learning for smart manufacturing: Methods and applications,

    J. Wang, Y . Ma, L. Zhang, R. X. Gao, and D. Wu, “Deep learning for smart manufacturing: Methods and applications,” Journal of manufac- turing systems, vol. 48, pp. 144–156, 2018

    work page 2018

  50. [50]

    Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems,

    A. D. Jagtap, E. Kharazmi, and G. E. Karniadakis, “Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems,”Computer Methods in Applied Mechanics and Engineering , vol. 365, p. 113028, 2020

    work page 2020

  51. [51]

    A high-efficient hybrid physics-informed neural networks based on convolutional neural network,

    Z. Fang, “A high-efficient hybrid physics-informed neural networks based on convolutional neural network,” IEEE Transactions on Neural Networks and Learning Systems , vol. 33, no. 10, pp. 5514–5526, 2021

    work page 2021

  52. [52]

    Physics informed neural network using finite difference method,

    K. L. Lim, R. Dutta, and M. Rotaru, “Physics informed neural network using finite difference method,” in 2022 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2022, pp. 1828–1833

    work page 2022

  53. [53]

    Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals,

    A. L. Goldberger, L. A. Amaral, L. Glass, J. M. Hausdorff, P. C. Ivanov, R. G. Mark, J. E. Mietus, G. B. Moody, C.-K. Peng, and H. E. Stanley, “Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals,” circulation, vol. 101, no. 23, pp. e215–e220, 2000

    work page 2000

  54. [54]

    Using inverse electrocardiography to image myocardial infarction—reflecting on the 2007 physionet/computers in cardiology challenge,

    F. Dawoud, G. S. Wagner, G. Moody, and B. M. Hor ´aˇcek, “Using inverse electrocardiography to image myocardial infarction—reflecting on the 2007 physionet/computers in cardiology challenge,” Journal of electrocardiology, vol. 41, no. 6, pp. 630–635, 2008

    work page 2007