Introduction to noninvasive cardiac mapping

Systematic review of computational techniques, dataset utilization, and feature extraction in electrocardiographic imaging

This study aimed to analyze computational techniques in ECG imaging (ECGI) reconstruction, focusing on dataset identification, problem-solving, and feature extraction. We employed a PRISMA approach to review studies from Scopus and Web of Science, applying Cochrane principles to assess risk of bias. The selection was limited to English peer-reviewed papers published from 2010 to 2023, excluding studies that lacked computational technique descriptions. From 99 reviewed papers, trends show a preference for traditional methods like the boundary element and Tikhonov methods, alongside a rising use of advanced technologies including hybrid techniques and deep learning. These advancements have enhanced cardiac diagnosis and treatment precision. Our findings underscore the need for robust data utilization and innovative computational integration in ECGI, highlighting promising areas for future research and advances. This shift toward tailored cardiac care suggests significant progress in diagnostic and treatment methods.

ECG electrode localization using 3D visual reconstruction

Body surface potential maps (BSPMs) derived from multi-channel ECG recordings enable the detection and diagnosis of electrophysiological phenomena beyond the standard 12-lead ECG. In this work, we developed two AI-based methods for the automatic detection of location of the electrodes used for BSPM: a rapid method using a specialized 3D Depth Sensing (DS) camera and a slower method that can use any 2D camera. Both methods were validated on a phantom model and in 7 healthy volunteers. With the phantom model, both 3D DS camera and 2D camera method achieved an average localization error less than 2 mm when compared to CT-scan or an Electromagnetic Tracking System (ETS). With healthy volunteers, the 3D camera yielded average 3D Euclidean distances ranging from 2.61 ± 1.2 mm to 5.78 ± 3.09 mm depending on the patient, similar to that seen with 2D camera (ranging from 2.45 ± 1.32 mm to 5.88 ± 2.73 mm). These results demonstrate high accuracy and provide practical alternatives to traditional imaging techniques, potentially enhancing the interest of BSPMs in a clinical setting.

Novel non-invasive ECG imaging method based on the 12-lead ECG for reconstruction of ventricular activation: A proof-of-concept study

Aim

Current non-invasive electrocardiographic imaging (ECGi) methods are often based on complex body surface potential mapping, limiting the clinical applicability. The aim of this pilot study was to evaluate the ability of a novel non-invasive ECGi method, based on the standard 12-lead ECG, to localize initial site of ventricular activation in right ventricular (RV) paced patients. Validation of the method was performed by comparing the ECGi reconstructed earliest site of activation against the true RV pacing site determined from cardiac computed tomography (CT).

Methods

This was a retrospective study using data from 34 patients, previously implanted with a dual chamber pacemaker due to advanced atrioventricular block. True RV lead position was determined from analysis of a post-implant cardiac CT scan. The ECGi method was based on an inverse-ECG algorithm applying electrophysiological rules. The algorithm integrated information from an RV paced 12-lead ECG together with a CT-derived patient-specific heart-thorax geometric model to reconstruct a 3D electrical ventricular activation map.

Results

The mean geodesic localization error (LE) between the ECGi reconstructed initial site of activation and the RV lead insertion site determined from CT was 13.9 ± 5.6 mm. The mean RV endocardial surface area was 146.0 ± 30.0 cm² and the mean circular LE area was 7.0 ± 5.2 cm² resulting in a relative LE of 5.0 ± 4.0%.

Conclusion

We demonstrated a novel non-invasive ECGi method, based on the 12-lead ECG, that accurately localized the RV pacing site in relation to the ventricular anatomy.

Tipping the scales of understanding: An engineering approach to design and implement whole-body cardiac electrophysiology experimental models

The study of cardiac electrophysiology is built on experimental models that span all scales, from ion channels to whole-body preparations. Novel discoveries made at each scale have contributed to our fundamental understanding of human cardiac electrophysiology, which informs clinicians as they detect, diagnose, and treat complex cardiac pathologies. This expert review describes an engineering approach to developing experimental models that is applicable across scales. The review also outlines how we applied the approach to create a set of multiscale whole-body experimental models of cardiac electrophysiology, models that are driving new insights into the response of the myocardium to acute ischemia. Specifically, we propose that researchers must address three critical requirements to develop an effective experimental model: 1) how the experimental model replicates and maintains human physiological conditions, 2) how the interventions possible with the experimental model capture human pathophysiology, and 3) what signals need to be measured, at which levels of resolution and fidelity, and what are the resulting requirements of the measurement system and the access to the organs of interest. We will discuss these requirements in the context of two examples of whole-body experimental models, a closed chest in situ model of cardiac ischemia and an isolated-heart, torso-tank preparation, both of which we have developed over decades and used to gather valuable insights from hundreds of experiments.

Solving Inverse Electrocardiographic Mapping Using Machine Learning and Deep Learning Frameworks

Electrocardiographic imaging (ECGi) reconstructs electrograms at the heart's surface using the potentials recorded at the body's surface. This is called the inverse problem of electrocardiography. This study aimed to improve on the current solution methods using machine learning and deep learning frameworks. Electrocardiograms were simultaneously recorded from pigs' ventricles and their body surfaces. The Fully Connected Neural network (FCN), Long Short-term Memory (LSTM), Convolutional Neural Network (CNN) methods were used for constructing the model. A method is developed to align the data across different pigs. We evaluated the method using leave-one-out cross-validation. For the best result, the overall median of the correlation coefficient of the predicted ECG wave was 0.74. This study demonstrated that a neural network can be used to solve the inverse problem of ECGi with relatively small datasets, with an accuracy compatible with current standard methods.

Electrocardiographic Imaging of Repolarization Abnormalities

Background

Dispersion and gradients in repolarization have been associated with life‐threatening arrhythmias, but are difficult to quantify precisely from surface electrocardiography. The objective of this study was to evaluate electrocardiographic imaging (ECGI) to noninvasively detect repolarization‐based abnormalities.

Methods and Results

Ex vivo data were obtained from Langendorff‐perfused pig hearts (n=8) and a human donor heart. Unipolar electrograms were recorded simultaneously during sinus rhythm from an epicardial sock and the torso‐shaped tank within which the heart was suspended. Regional repolarization heterogeneities were introduced through perfusion of dofetilide and pinacidil into separate perfusion beds. In vivo data included torso and epicardial potentials recorded simultaneously in anesthetized, closed‐chest pigs (n=5), during sinus rhythm, and ventricular pacing. For both data sets, ECGI accurately reconstructed T‐wave electrogram morphologies when compared with those recorded by the sock (ex vivo: correlation coefficient, 0.85 [0.52–0.96], in vivo: correlation coefficient, 0.86 [0.52–0.96]) and repolarization time maps (ex‐vivo: correlation coefficient, 0.73 [0.63–0.83], in vivo: correlation coefficient, 0.76 [0.67–0.82]). ECGI‐reconstructed repolarization time distributions were strongly correlated to those measured by the sock (both data sets, R² ≥0.92). Although the position of the gradient was slightly shifted by 8.3 (0–13.9) mm, the mean, max, and SD between ECGI and recorded gradient values were highly correlated (R²=0.87, 0.75, and 0.86 respectively). There was no significant difference in ECGI accuracy between ex vivo and in vivo data.

Conclusions

ECGI reliably and accurately maps potentially critical repolarization abnormalities. This noninvasive approach allows imaging and quantifying individual parameters of abnormal repolarization‐based substrates in patients with arrhythmogenesis, to improve diagnosis and risk stratification.

Novel experimental model for studying the spatiotemporal electrical signature of acute myocardial ischemia: a translational platform

Myocardial ischemia is one of the most common cardiovascular pathologies and can indicate many severe and life threatening diseases. Despite these risks, current electrocardiographic detection techniques for ischemia are mediocre at best, with reported sensitivity and specificity ranging from 50%-70% and 70%-90%, respectively.

OBJECTIVE

To improve this performance, we set out to develop an experimental preparation to induce, detect, and analyze bioelectric sources of myocardial ischemia and determine how these sources reflect changes in body-surface potential measurements.

APPROACH

We designed the experimental preparation with three important characteristics: (1) enable comprehensive and simultaneous high-resolution electrical recordings within the myocardial wall, on the heart surface, and on the torso surface; (2) develop techniques to visualize these recorded electrical signals in time and space; and (3) accurately and controllably simulate ischemic stress within the heart by modulating the supply of blood, the demand for perfusion, or a combination of both.

MAIN RESULTS

To achieve these goals we designed comprehensive system that includes (1) custom electrode arrays (2) signal acquisition and multiplexing units, (3) a surgical technique to place electrical recording and myocardial ischemic control equipment, and (4) an image based modeling pipeline to acquire, process, and visualize the results. With this setup, we are uniquely able to capture simultaneously and continuously the electrical signatures of acute myocardial ischemia within the heart, on the heart surface, and on the body surface.

SIGNIFICANCE

This novel experimental preparation enables investigation of the complex and dynamic nature of acute myocardial ischemia that should lead to new, clinically translatable results.

Inverse estimation of cardiac activation times via gradient-based optimization

Computational modeling may provide a quantitative framework for integrating multiscale data to gain insight into mechanisms of heart disease, identify and test pharmacological and electrical therapy and interventions, and support clinical decisions. Patient-specific computational cardiac models can help guide such procedures, and cardiac inverse modeling is a promising alternative to adequately personalize these models. Indeed, full cardiac inverse modeling is currently becoming computationally feasible; however, fundamental work to assess the feasibility of emerging techniques is still needed. In this study, we use a partial differential equation-constrained optimal control approach to numerically investigate the identifiability of an initial activation sequence from synthetic (partial) observations of the extracellular potential using the bidomain approximation and 2D representations of cardiac tissue. Our results demonstrate that activation times and duration of several stimuli can be recovered even with high levels of noise, that it is sufficient to sample the observations at the electrocardiogram-relevant sampling frequency of 1 kHz, and that spatial resolutions that are coarser than the standard in electrophysiological simulations can be used. The optimization of activation times is still effective when synthetic data are generated with a different cell membrane kinetics model than optimized for. The findings thus indicate that the presented approach has potential for finding activation sequences from clinical data modalities, as an extension to existing cardiac imaging approaches.

Experimental Data and Geometric Analysis Repository-EDGAR

INTRODUCTION

The "Experimental Data and Geometric Analysis Repository", or EDGAR is an Internet-based archive of curated data that are freely distributed to the international research community for the application and validation of electrocardiographic imaging (ECGI) techniques. The EDGAR project is a collaborative effort by the Consortium for ECG Imaging (CEI, ecg-imaging.org), and focused on two specific aims. One aim is to host an online repository that provides access to a wide spectrum of data, and the second aim is to provide a standard information format for the exchange of these diverse datasets.

METHODS

The EDGAR system is composed of two interrelated components: 1) a metadata model, which includes a set of descriptive parameters and information, time signals from both the cardiac source and body-surface, and extensive geometric information, including images, geometric models, and measure locations used during the data acquisition/generation; and 2) a web interface. This web interface provides efficient, search, browsing, and retrieval of data from the repository.

RESULTS

An aggregation of experimental, clinical and simulation data from various centers is being made available through the EDGAR project including experimental data from animal studies provided by the University of Utah (USA), clinical data from multiple human subjects provided by the Charles University Hospital (Czech Republic), and computer simulation data provided by the Karlsruhe Institute of Technology (Germany).

CONCLUSIONS

It is our hope that EDGAR will serve as a communal forum for sharing and distribution of cardiac electrophysiology data and geometric models for use in ECGI research.