MACHINE LEARNING BASED RECONSTRUCTION OF INTRACARDIAC ELECTRICAL BEHAVIOR BASED ON ELECTROCARDIOGRAMS
A computer-based system and process are disclosed for reconstructing the internal electrical behavior of a patient's heart based partly or wholly on the patient's electrocardiogram (ECG). The output of the process may include, for example, a cardiac activation map, and/or a representation of transmembrane potentials over time. The process advantageously does not require any medical imaging of the patient, and does not require any special medical equipment. For example, the patient's activation map and transmembrane potentials may be reconstructed based solely on a preexisting or newly-obtained 12-lead cardiac ECG of the patient. The process makes use of a machine learning model, such as a neural network based model, trained with actual and/or simulated ECGs and intracardiac electrical data (typically transmembrane potentials) of many thousands of patients. Because an insufficient quantity of such data exists for actual patients, model training may be performed using ECGs and intracardiac electrical data obtained through computer simulations.
1 . A process for generating a machine learning model capable of predicting intracardiac electrical behavior of a patient using non-invasively acquired data, the process comprising:
generating simulated electrocardiograms (ECGs) and simulated intracardiac electrical behavior data for each of a plurality of cardiac geometries; and
training a machine learning model to predict intracardiac electrical behavior based on ECG characteristics, wherein training the machine learning model comprises performing feature extraction on the simulated ECGs and simulated intracardiac electrical behavior data to identify features of the simulated ECGs and features of corresponding intracardiac electrical behavior, and generating model weights that represent correlations between the features of the simulated ECGs and features of the corresponding intracardiac electrical behavior;
said process performed by a computing system comprising one or more computing devices.
2 . The process of claim 1 , wherein extracting features of the simulated ECGs comprises performing wavelet decomposition of the ECGs.
3 . The process of claim 1 , further comprising performing validation of the trained machine learning model using actual ECG and corresponding actual intracardiac electrical behavior of patients.
4 . The process of claim 1 , further comprising using the trained machine learning model to predict intracardiac electrical behavior of a patient based on an ECG of the patient.
5 . The process of claim 1 , wherein the machine learning model comprises a sequence-to-sequence machine learning model.
6 . The process of claim 5 , wherein the sequence-to-sequence machine learning model comprises a sequence-to-sequence neural network.
7 . Non-transitory computer storage having stored thereon a trained machine learning model capable of predicting intracardiac electrical behavior of a patient using non-invasively acquired data, the trained machine learning model generated by a process that comprises:
generating simulated electrocardiograms (ECGs) and simulated intracardiac electrical behavior data for each of a plurality of cardiac geometries; and
training a machine learning model to predict intracardiac electrical behavior based on ECG characteristics, wherein training the machine learning model comprises performing feature extraction on the simulated ECGs and simulated intracardiac electrical behavior data to identify features of the simulated ECGs and features of corresponding intracardiac electrical behavior, and generating model weights that represent correlations between the features of the simulated ECGs and features of the corresponding intracardiac electrical behavior.
8 . The non-transitory computer storage of claim 7 , wherein extracting features of the simulated ECGs comprises performing wavelet decomposition of the simulated ECGs.
9 . The non-transitory computer storage of claim 7 , wherein the process further comprises performing validation of the trained machine learning model using actual ECG and corresponding actual intracardiac electrical behavior of patients.
10 . The non-transitory computer storage of claim 7 , wherein the machine learning model comprises a sequence-to-sequence machine learning model.
11 . The non-transitory computer storage of claim 10 , wherein the sequence-to-sequence machine learning model comprises a sequence-to-sequence neural network.