C. Nagel, M. Schaufelberger, O. Dössel, and A. Loewe. A Bi-atrial Statistical Shape Model as a Basis to Classify Left Atrial Enlargement from Simulated and Clinical 12-Lead ECGs. In Statistical Atlases and Computational Models of the Heart. Multi-Disease, Multi-View, and Multi-Center Right Ventricular Segmentation in Cardiac MRI Challenge, vol. 13131, pp. 38-47, 2022
Left atrial enlargement (LAE) is one of the risk factors for atrial fibrillation (AF). A non-invasive and automated detection of LAE with the 12-lead electrocardiogram (ECG) could therefore contribute to an improved AF risk stratification and an early detection of new-onset AF incidents. However, one major challenge when applying machine learning techniques to identify and classify cardiac diseases usually lies in the lack of large, reliably labeled and balanced clinical datasets. We therefore examined if the extension of clinical training data by simulated ECGs derived from a novel bi-atrial shape model could improve the automated detection of LAE based on P waves of the 12-lead ECG. We derived 95 volumetric geometries from the bi-atrial statistical shape model with continuously increasing left atrial volumes in the range of 30 ml to 65 ml. Electrophysiological simulations with 10 different conduction velocity settings and 2 different torso models were conducted. Extracting the P waves of the 12-lead ECG thus yielded a synthetic dataset of 1,900 signals. Besides the simulated data, 7,168 healthy and 309 LAE ECGs from a public clinical ECG database were available for training and testing of an LSTM network to identify LAE. The class imbalance of the training data could be reduced from 1:23 to 1:6 when adding simulated data to the training set. The accuracy evaluated on the test dataset comprising a subset of the clinical ECG recordings improved from 0.91 to 0.95 if simulated ECGs were included as an additional input for the training of the classifier. Our results suggest that using a bi-atrial statistical shape model as a basis for ECG simulations can help to overcome the drawbacks of clinical ECG recordings and can thus lead to an improved performance of machine learning classifiers to detect LAE based on the 12-lead ECG.
The bidomain model and the finite element method are an established standard to mathematically describe cardiac electrophysiology, but are both suboptimal choices for fast and large-scale simulations due to high computational costs. We investigate to what extent simplified approaches for propagation models (monodomain, reaction-Eikonal and Eikonal) and forward calculation (boundary element and infinite volume conductor) deliver markedly accelerated, yet physiologically accurate simulation results in atrial electrophysiology. <i>Methods:</i> We compared action potential durations, local activation times (LATs), and electrocardiograms (ECGs) for sinus rhythm simulations on healthy and fibrotically infiltrated atrial models. <i>Results:</i> All simplified model solutions yielded LATs and P waves in accurate accordance with the bidomain results. Only for the Eikonal model with pre-computed action potential templates shifted in time to derive transmembrane voltages, repolarization behavior notably deviated from the bidomain results. ECGs calculated with the boundary element method were characterized by correlation coefficients <inline-formula><tex-math notation="LaTeX">$>$</tex-math></inline-formula>0.9 compared to the finite element method. The infinite volume conductor method led to lower correlation coefficients caused predominantly by systematic overestimations of P wave amplitudes in the precordial leads. <i>Conclusion:</i> Our results demonstrate that the Eikonal model yields accurate LATs and combined with the boundary element method precise ECGs compared to markedly more expensive full bidomain simulations. However, for an accurate representation of atrial repolarization dynamics, diffusion terms must be accounted for in simplified models. <i>Significance:</i> Simulations of atrial LATs and ECGs can be notably accelerated to clinically feasible time frames at high accuracy by resorting to the Eikonal and boundary element methods.
Mechanistic cardiac electrophysiology models allow for personalized simulations of the electrical activity in the heart and the ensuing electrocardiogram (ECG) on the body surface. As such, synthetic signals possess known ground truth labels of the underlying disease and can be employed for validation of machine learning ECG analysis tools in addition to clinical signals. Recently, synthetic ECGs were used to enrich sparse clinical data or even replace them completely during training leading to improved performance on real-world clinical test data. We thus generated a novel synthetic database comprising a total of 16,900 12 lead ECGs based on electrophysiological simulations equally distributed into healthy control and 7 pathology classes. The pathological case of myocardial infraction had 6 sub-classes. A comparison of extracted features between the virtual cohort and a publicly available clinical ECG database demonstrated that the synthetic signals represent clinical ECGs for healthy and pathological subpopulations with high fidelity. The ECG database is split into training, validation, and test folds for development and objective assessment of novel machine learning algorithms.
Interatrial conduction block refers to a disturbance in the propagation of electrical impulses in the conduction pathways between the right and the left atrium. It is a risk factor for atrial fibrillation, stroke, and premature death. Clin- ical diagnostic criteria comprise an increased P wave dura- tion and biphasic P waves in lead II, III and aVF due to ret- rograde activation of the left atrium. Machine learning algo- rithms could improve the diagnosis but require a large-scale, well-controlled and balanced dataset. In silico electrocardio- gram (ECG) signals, optimally obtained from a statistical shape model to cover anatomical variability, carry the poten- tial to produce an extensive database meeting the requirements for successful machine learning application. We generated the first in silico dataset including interatrial conduction block of 9,800simulated ECG signals based on a bi-atrial statistical shape model. Automated feature analysis was performed to evaluate P wave morphology, duration and P wave terminal force in lead V1. Increased P wave duration and P wave ter- minal force in lead V1 were found for models with interatrial conduction block compared to healthy models. A wide vari- ability of P wave morphology was detected for models with in- teratrial conduction block. Contrary to previous assumptions, our results suggest that a biphasic P wave morphology seems to be neither necessary nor sufficient for the diagnosis of in- teratrial conduction block. The presented dataset is ready for a classification with machine learning algorithms and can be easily extended.
Aims Atrial flutter (AFlut) is a common re-entrant atrial tachycardia driven by self-sustainable mechanisms that cause excitations to propagate along pathways different from sinus rhythm. Intra-cardiac electrophysiological mapping and catheter ablation are often performed without detailed prior knowledge of the mechanism perpetuating AFlut, likely prolonging the procedure time of these invasive interventions. We sought to discriminate the AFlut location [cavotricuspid isthmus-dependent (CTI), peri-mitral, and other left atrium (LA) AFlut classes] with a machine learning-based algorithm using only the non-invasive signals from the 12-lead electrocardiogram (ECG). Methods and results Hybrid 12-lead ECG dataset of 1769 signals was used (1424 in silico ECGs, and 345 clinical ECGs from 115 patients—three different ECG segments over time were extracted from each patient corresponding to single AFlut cycles). Seventy-seven features were extracted. A decision tree classifier with a hold-out classification approach was trained, validated, and tested on the dataset randomly split after selecting the most informative features. The clinical test set comprised 38 patients (114 clinical ECGs). The classifier yielded 76.3% accuracy on the clinical test set with a sensitivity of 89.7%, 75.0%, and 64.1% and a positive predictive value of 71.4%, 75.0%, and 86.2% for CTI, peri-mitral, and other LA class, respectively. Considering majority vote of the three segments taken from each patient, the CTI class was correctly classified at 92%. Conclusion Our results show that a machine learning classifier relying only on non-invasive signals can potentially identify the location of AFlut mechanisms. This method could aid in planning and tailoring patient-specific AFlut treatments.
C. Nagel. Multiscale Cohort Modeling of Atrial Electrophysiology : Risk Stratification for Atrial Fibrillation through Machine Learning on Electrocardiograms. Karlsruher Institut für Technologie (KIT). Dissertation. 2022
Patients affected by atrial fibrillation are exposed to a fivefold increased risk of ischemic stroke. An early detection and diagnosis of the arrhythmia would therefore set the course for timely intervention to prevent potentially occurring comorbidities. A dilation of the left atrium as well as the presence of fibrotically infiltrated atrial tissue are risk factors for atrial fibrillation as they provide the necessary substrate for the maintenance of electrical reentrant activity. Identifying fibrotic atrial cardiomyopathy and left atrial enlargement based on machine learning techniques applied to P waves representing the atrial activity in the 12-lead electrocardiogram in sinus rhythm could thus be an important means for a non-invasive and remote risk stratification of new-onset atrial fibrillation episodes and the selection of appropriate subjects for in-depth screening.For this purpose, it was investigated if simulated atrial electrocardiogram data added to a clinical training dataset of a machine learning model could contribute to an improved classification of the above mentioned diseases. Two virtual cohorts characterized by both anatomical and functional variability were compiled and served as a basis for generating large-scale and unbiased datasets of P waves with exactly known ground truth labels of the underlying pathology. In this way, the simulated data fulfilled the essential requirements for the development of a machine learning algorithm what sets them apart from clinical data usually not available in large numbers in evenly distributed classes and labels possibly affected by inadequate expert annotations.A shallow feature-based feedforward neural network was set up to perform the regression task of predicting the tissue volume fraction of left atrial fibrosis. Compared to training the model only on clinical data, training on a hybrid dataset led to a reduction of the absolute estimation error from 17.5 % fibrotic volume on average to 16.5 % evaluated on a clinical test set. A long short-term memory network tailored at performing the binary classification task between P waves of healthy subjects and left atrial enlargement patients yielded an accuracy on a clinical test set of 0.95 when trained on a hybrid dataset, of 0.91 when trained on clinical data only comprising samples with 100 % label certainties and of 0.83 when trained on clinical data including all samples independent on their label certainties.The results of the studies presented in this thesis demonstrate that electrocardiogram data resulting from electrophysiological modeling and simulations on virtual patient cohorts, covering relevant variability aspects complying with real-world observations, can be a valuable data resource for improving automated atrial fibrillation risk stratification. In this regard, the drawbacks of clinical datasets for developing machine learning models can be compensated for. This ultimately leads to an enhanced early detection of the arrhythmia, which allows for choosing appropriate treatment strategies in due time and thus, reduces the risk of stroke in affected patients.