Automatic fibrillation classification and identification of fibrillation epochs
Methods and computer systems are described that classify a cardiogram as being an atrial fibrillation (AF) or ventricular fibrillation (VF) cardiogram, automatically detect an AF epoch within an AF cardiogram, and automatically detect a VF epoch within a VF cardiogram. A classification and identification (C&I) system includes a classification system, an AF identification system, and a VF identification system. The C&I system processes cardiograms collected from patients to classify the cardiograms as being AF cardiograms or VF cardiograms and to identify AF epochs within the AF cardiograms or VF epochs within the VF cardiograms. The C&I system may then identify an AF source location of an AF based on the AF epochs and a VF source location of a VF based on the VF epochs. The C&I system may display a graphic of a heart that includes an indication of a source location.
1. One or more computing systems for identifying an atrial fibrillation (AF) epoch of a patient based at least on a patient cardiogram collected from a patient, the one or more computing systems comprising:
one or more computer-readable storage mediums that store:
training T-Q intervals of cardiograms that are AF epochs or not AF epochs; and
computer-executable instructions for controlling the one or more computing systems to:
generate training data that includes, for each of a plurality of the training T-Q intervals, a feature vector derived from that training T-Q interval that is labeled as an AF epoch or not an AF epoch;
train an AF epoch machine learning (ML) model using the generated training data, the AF epoch ML model being a neural network having connected layers with activation functions having weights, wherein the AF epoch ML model is trained to learn weights of the activation functions that tend to minimize a loss function;
apply an AF classification ML model to the patient cardiogram to determine whether the patient cardiogram is an AF cardiogram; and
when the patient cardiogram is an AF cardiogram,
identify one or more landmarks within the patient cardiogram;
identify a T-Q interval within the patient cardiogram based on the identified one or more landmarks;
apply the AF epoch ML model to a feature vector derived from the identified T-Q interval to determine whether the identified T-Q interval represents an AF epoch;
apply a mapping system to the identified T-Q interval to identify a source location of the AF;
output the source location of the AF to an ablation therapy device; and
direct the ablation therapy device to perform an ablation targeting the source location; and
one or more processors for controlling the one or more computing systems to execute one or more of the computer-executable instructions.
2. The one or more computing systems of claim 1 wherein the computer-executable instructions further include instructions that:
display a graphic of a heart that illustrates the source location of the AF.
3. The one or more computing systems of claim 1 wherein the computer-executable instructions that identify one or more landmarks identify one or more T peaks and Q peaks.
4. The one or more computing systems of claim 1 wherein the training T-Q intervals include simulated training T-Q intervals generated based on simulations of electric activity of hearts having different characteristics.
5. The one or more computing systems of claim 1 wherein the training T-Q intervals include clinical training T-Q intervals collected from patients.
6. The one or more computing systems of claim 1 wherein the computer-executable instructions further comprise instructions to identify AF cycles within an AF epoch and filter out one or more AF cycles that do not satisfy an AF cycle criterion.
7. The one or more computing systems of claim 1 wherein the feature vector includes an image of a T-Q interval and the AF epoch ML model includes a convolutional neural network.
8. The one or more computing systems of claim 1 wherein the AF epoch ML model includes a transformer.
9. The one or more computing systems of claim 1 wherein the feature vector includes an image of a T-Q interval and the AF epoch ML model includes a generative adversarial network.
10. The one or more computing systems of claim 1 wherein the feature vector includes a voltage-time series representation of a T-Q interval and the AF epoch ML model includes a recurrent neural network.
11. A method for treating a patient, the method comprising:
performing by one or more computing systems identification of an atrial fibrillation (AF) epoch of the patient, the method comprising:
accessing a patient cardiogram collected from the patient;
applying an AF classification machine learning (ML) model to the patient cardiogram to determine whether the patient cardiogram is an AF cardiogram, the AF classification ML model being a neural network that is trained using classification training data that includes feature vectors representing one or more features of training cardiograms and labels indicating whether the training cardiograms are AF cardiograms; and
when the patient cardiogram is an AF cardiogram,
identifying one or more landmarks within the patient cardiogram;
identifying a T-Q interval within the patient cardiogram based on the identified one or more landmarks; and
applying an AF epoch ML model to a feature vector with one or more features derived from the identified T-Q interval to determine whether the identified T-Q interval represents an AF epoch, the AF epoch ML model trained using training data that includes, for each of a plurality of the training T-Q intervals, a feature vector derived from that training T-Q interval that is labeled as an AF epoch or not an AF epoch; and
performing an ablation on the patient based on the determination that the identified T-Q interval represents an AF epoch.
12. The method of claim 11 further comprising training the AF epoch ML model using the epoch training data.
13. The method of claim 11 further comprising when the identified T-Q interval represents an AF epoch,
applying a mapping system to the identified T-Q interval to identify a source location of the AF represented by the AF epoch; and
displaying a graphic of a heart that illustrates the source location of the AF.
14. The method of claim 11 further comprising when the identified T-Q interval represents an AF epoch,
applying a mapping system to the identified T-Q interval to identify a source location of the AF represented by the AF epoch; and
outputting the source location of the AF to an ablation therapy device to control the performing of the ablation targeting the source location.
15. A method for treating a target patient, the method comprising:
performing by one or more computing system for training of an atrial fibrillation (AF) epoch machine learning (ML) model, the method comprising:
accessing a plurality of training T-Q intervals along with indications of whether the training T-Q intervals represent an AF epoch or not an AF epoch;
for each training T-Q interval, generating training data that includes a feature vector with one or more features derived from that training T-Q interval and a label indicating whether that training T-Q interval represents an AF epoch or not an AF epoch;
training the AF epoch ML model that is a neural network using the training data wherein the AF epoch ML model inputs a feature vector with one or more features derived from a T-Q interval of a patient cardiogram and outputs an indication of whether the T-Q interval represents an AF epoch; and
inputting to the trained AF epoch ML model features derived from a target patient T-Q interval of a target patient cardiogram of the target patient wherein the trained AF epoch ML model outputs an indication of whether the target patient T-Q interval represents an AF each; and
performing an ablation on the target patient based on the target patient T-Q interval representing an AF epoch.
16. The method of claim 15 further comprising identifying a training T-Q intervals from a training cardiogram by identifying one or more landmarks that include one or more T peaks and one or more Q peaks.
17. The method of claim 15 wherein the training T-Q intervals include simulated training T-Q intervals generated based on simulations of electric activity of hearts having different characteristics and/or clinical training T-Q intervals collected from patients.
18. The method of claim 15 wherein the feature vector includes an image of a T-Q interval and the AF epoch ML model includes a convolutional neural network.
19. The method of claim 15 wherein the AF epoch ML model includes a transformer.
20. The method of claim 15 wherein the feature vector includes an image of a T-Q interval and the AF epoch ML model includes a generative adversarial network.
21. The method of claim 15 wherein the feature vector includes a voltage-time series representation of a T-Q interval and the AF epoch ML model includes a recurrent neural network.
22. The method of claim 15 wherein a feature vector includes a feature that is an image of a T-Q interval and the AF epoch ML model includes a convolutional neural network.
23. One or more computing systems for identifying an atrial fibrillation (AF) epoch of a patient cardiogram collected from a patient, the one or more computing systems comprising:
one or more computer-readable storage mediums that store computer-executable instructions for controlling the one or more computing systems to:
receive patient cardiogram that was collected from the patient, the patient cardiogram representing an AF cardiogram;
identify one or more landmarks within the patient cardiogram;
identify a T-Q interval within the patient cardiogram based on the identified one or more landmarks;
apply an AF epoch machine learning (ML) model to a feature vector with one or more features derived from the identified T-Q interval to determine whether the identified T-Q interval represents an AF epoch, the AF epoch ML model trained using training data that includes, for each of a plurality of the training T-Q intervals, a feature vector derived from that training T-Q interval that is labeled as an AF epoch or not an AF epoch; and
apply a mapping system to the identified T-Q interval to identify a source location of the AF;
control an ablation therapy device to perform an ablation targeting the source location; and
one or more processors for controlling the one or more computing systems to execute one or more of the computer-executable instructions.
24. The one or more computing systems of claim 23 wherein the one or more of the computing systems is a cloud-based system and the patient cardiogram is received from a client-side computing system and the indication of the determination is output to the client-side computing system.
25. The one or more computing systems of claim 23 wherein the computer-executable instruction include computer-executable instructions that control the one or more computing systems to train the AF epoch ML model using the training data.
26. The one or more computing systems of claim 23 wherein the computer-executable instruction include computer-executable instructions that control the one or more computing systems to apply a mapping system to the identified T-Q interval to identify a source location of the AF represented by the AF epoch and to display a graphic of a heart that illustrates the source location of the AF.
27. The one or more computing systems of claim 23 wherein the AF epoch ML model includes a transformer.
28. The one or more computing systems of claim 23 wherein a feature vector a feature that is an image of a T-Q interval and the AF epoch ML model includes a generative adversarial network.