Noninvasive Reconstruction of Transmural Transmembrane Potential With Simultaneous Estimation of Prior Model Error

Q: What is the clinical evidence from this study?

Population: Post-infarction ventricular arrhythmia (n=2). Intervention: Bayesian noninvasive electrophysiological imaging with simultaneous estimation of prior model error vs. Model-constrained transmembrane potential (TMP) reconstruction without correcting prior model errors. Primary outcome: Accuracy of 3D transmural transmembrane potential (TMP) reconstruction and myocardial infarct detection (evaluated via correlation coefficient for activation time, normalized mean square error, and Dice coefficient for infarct region).

Q: What does this research mean for the field?

Simultaneously estimating prior model errors using a sparse gradient prior in Bayesian inference significantly improves the accuracy of noninvasive 3D transmural transmembrane potential reconstruction and myocardial infarct localization. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Sandesh Ghimire; John L. Sapp; B. Milan Horáček; Linwei Wang

doi:10.1109/tmi.2019.2906600

March 20, 2019IEEE Transactions on Medical ImagingOpen Access

Noninvasive Reconstruction of Transmural Transmembrane Potential With Simultaneous Estimation of Prior Model Error

View Full Paper

Ask AI

Bookmark

Share

Key result

Bayesian error-corrected imaging improves noninvasive 3D transmural potential reconstruction and infarct localization.

n=2

Why the study?

Electrophysiological models used to regularize inverse ECG reconstruction problems suffer from errors related to tissue properties and earliest excitation sites.

Does simultaneous Bayesian estimation of sparse prior model error improve the accuracy of noninvasive 3D transmural transmembrane potential reconstruction and infarct localization from 120-lead ECGs compared to standard model-constrained reconstruction without error correction?

Key Points

To develop and evaluate a Bayesian framework that simultaneously reconstructs transmural transmembrane potentials and estimates prior electrophysiological model errors from body-surface electrocardiograms.
Formulated a Bayesian framework that enforces sparsity of prior model errors in the gradient domain via a variational lower bound of the generalized Gaussian distribution.
Evaluated the method on three image-derived heart-torso models across 34 myocardial infarct configurations and initial excitation site errors using simulated 120-lead ECG data with 20dB noise.
Tested the algorithm on clinical data from two patients undergoing catheter ablation for post-infarction ventricular arrhythmia, comparing reconstructed infarcts to in-vivo bipolar voltage maps (≤ 1.5 mV).
Simultaneous error estimation detected conduction blocks early in depolarization and yielded higher activation time correlation and infarct detection Dice coefficients, particularly for non-septal infarcts.
Corrected initial condition discrepancies, including identifying an omitted excitation point within 20ms into depolarization.
Demonstrated higher Dice coefficients and closer visual alignment to in-vivo bipolar voltage maps across both clinical cases compared to reconstructions without model error correction.

Structured PICO

Does simultaneous Bayesian estimation of sparse prior model error improve the accuracy of noninvasive 3D transmural transmembrane potential reconstruction and infarct localization from 120-lead ECGs compared to standard model-constrained reconstruction without error correction?

P

Population

3 image-derived synthetic heart-torso models (34 infarct settings) and 2 clinical patients undergoing catheter ablation for post-infarction ventricular arrhythmia.

I

Intervention

Bayesian noninvasive electrophysiological (EP) imaging algorithm that simultaneously reconstructs 3D transmural transmembrane potential (TMP) from 120-lead body-surface ECGs and estimates prior Aliev-Panfilov EP model error using a sparse prior in the spatial gradient domain (based on a variational lower bound of the generalized Gaussian distribution).

C

Comparator

Previously-described model-constrained 3D transmural TMP reconstruction approach (using the Aliev-Panfilov model) that does not consider or correct errors in the a priori physiological model.

O

Outcome

Quantitative accuracy of reconstructed 3D transmural TMP sequences and infarct localization, measured by correlation coefficient (activation time and TMP sequence), normalized mean square error (TMP sequence), and Dice coefficient (between true/low-voltage ≤1.5 mV bipolar map infarct regions and reconstructed regions where TMP duration falls below 50% of normal).surrogate

Incorporating sparse Bayesian estimation of prior physiological model errors significantly improves noninvasive 3D cardiac electrical imaging of myocardial infarction and ectopic activation from surface ECGs, though clinical translation remains challenged by forward-model inaccuracies and poor septal observability.

Limitations

In real-data experiments, errors in the forward measurement matrix H directly affect the accuracy of inverse reconstructions.
If the error between true TMP propagation and the prior EP model is too high (e.g., complex realistic infarct geometries or unpredictable excitation points), the assumption of sparse model error may not hold.
Correspondence between invasive bipolar voltage maps and reconstructed TMP duration is indirect, and registration of voltage maps to CT-derived meshes introduces additional validation errors.
Ability to reconstruct septal infarcts is fundamentally limited by poor observability on surface ECG data.
The method performs inference sequentially using only past ECG data because the prior EP model cannot be reversed in time.
Errors in the forward matrix H in real-data experiments degrade inverse reconstruction accuracy
Assumption of sparse model error may not hold when complex infarct distributions or unpredictable excitation points cause large deviations from the prior EP model
Indirect correspondence and mesh registration errors between invasive bipolar voltage maps and CT-derived TMP reconstructions
Poor reconstruction performance for septal infarcts due to limited observability in body-surface ECG data
Sequential inference only uses past ECG data because the prior EP simulation model cannot be reversed in time

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May enhance noninvasive ECG-based infarct imaging accuracy; leaves open prospective clinical validation before adoption.

Abstract

To reconstruct electrical activity in the heart from body-surface electrocardiograms (ECGs) is an ill-posed inverse problem. Electrophysiological models have been found effective in regularizing these inverse problems by incorporating a priori knowledge about how the electrical potential in the heart propagates over time. However, these models suffer from model errors arising from, for example, parameters associated with tissue properties and the earliest sites of excitation. We present a Bayesian approach to simultaneously estimate transmembrane potential (TMP) signals and prior model errors, exploiting sparsity of the error in the gradient domain in the form of a novel sparse prior based on variational lower bound of the generalized Gaussian distribution. In synthetic and real-data experiments, we demonstrate the improvement of accuracy in TMP reconstruction brought by simultaneous model error estimation. We further provide theoretical and empirical justifications for the change of performances in the presented method at the presence of different model errors.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Ghimire et al. (2019) studied Post-infarction ventricular arrhythmia (n=2). Bayesian noninvasive electrophysiological imaging with simultaneous estimation of prior model error vs. Model-constrained transmembrane potential (TMP) reconstruction without correcting prior model errors was evaluated on Accuracy of 3D transmural transmembrane potential (TMP) reconstruction and myocardial infarct detection (evaluated via correlation coefficient for activation time, normalized mean square error, and Dice coefficient for infarct region). Bayesian inference with simultaneous sparse prior model error estimation significantly improved noninvasive 3D transmural transmembrane potential reconstruction and infarct localization compared to uncorrected reconstruction.

synapsesocial.com/papers/6a79209a1ad0da161567566a https://doi.org/10.1109/tmi.2019.2906600

View Full Paper

Ask AI

Bookmark

Share

Noninvasive Reconstruction of Transmural Transmembrane Potential With Simultaneous Estimation of Prior Model Error

Key Points

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Noninvasive Reconstruction of Transmural Transmembrane Potential With Simultaneous Estimation of Prior Model Error

Key Points

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading