Machine learning using clinical and simulated data
Systems are provided for generating data representing electromagnetic states of a heart for medical, scientific, research, and/or engineering purposes. The systems generate the data based on source configurations such as dimensions of, and scar or fibrosis or pro-arrhythmic substrate location within, a heart and a computational model of the electromagnetic output of the heart. The systems may dynamically generate the source configurations to provide representative source configurations that may be found in a population. For each source configuration of the electromagnetic source, the systems run a simulation of the functioning of the heart to generate modeled electromagnetic output (e.g., an electromagnetic mesh for each simulation step with a voltage at each point of the electromagnetic mesh) for that source configuration. The systems may generate a cardiogram for each source configuration from the modeled electromagnetic output of that source configuration for use in predicting the source location of an arrhythmia.
1. One or more computing systems for generating a patient classifier for classifying a cardiogram, the one or more computing systems comprising:
one or more computer-readable storage mediums storing computer-executable instructions for controlling the one or more computing systems to:
for each of a plurality of persons,
access a cardiogram collected from that person and a classification for that person; and
generate a feature vector that includes the cardiogram and a label based on the classification for that person; and
train the patient classifier with training data that includes the feature vectors and the labels, the patient classifier for inputting a patient cardiogram of a patient and outputting a classification for that patient; and
one or more processors for executing the computer-executable instructions stored in the one or more computer-readable storage mediums.
2. The one or more computing systems of claim 1 wherein the patient classifier is a convolutional neural network.
3. The one or more computing systems of claim 1 wherein the patient classifier is a neural network.
4. The one or more computing systems of claim 1 wherein the training data further includes, for each of a plurality of simulations, a feature vector that includes a modeled cardiogram generated based on the simulation and a label that is based on a classification used in the simulation.
5. The one or more computing systems of claim 1 wherein the computer-executable instructions further control the one or more computing systems to train a model classifier using modeled training data the includes, for each of a plurality of simulations, a modeled cardiogram generated based on the simulation and a label that is based on a classification used in the simulation and wherein weights of the patient classifier are initialized based on weights learned during the simulations.
6. The one or more computing systems of claim 1 wherein the computer-executable instructions further control the one or more computing systems to, for each of a plurality of clusters of similar persons, train a cluster patient classifier based on training data that includes cardiograms and classifications for the persons in that cluster.
7. The one or more computing systems of claim 6 wherein the computer-executable instructions further control the one or more computing systems to identify similar persons based on comparison of heart configurations of the persons.
8. The one or more computing systems of claim 6 wherein the computer-executable instructions further control the one or more computing systems to identify similar persons based on comparison of cardiograms of the persons.
9. The one or more computing systems of claim 8 wherein the computer-executable instructions further control the one or more computing systems to:
identify a cluster of similar persons that are similar to a target patient; and
apply the cluster patient classifier for that identified cluster to a target patient cardiogram of the target patient to generate a target patient classification for the target patient.
10. One or more computing systems for performing machine learning training, the one or more computing systems comprising:
one or more computer-readable storage mediums storing computer-executable instructions for controlling the one or more computing systems to:
access a plurality of training cardiograms, each of the training cardiograms derived from electromagnetic output associated with a heart and associated with a value for a configuration parameter of a source configuration of the heart, the electromagnetic output representing electrical activity of the heart; and
performing machine learning training based on the training cardiograms to generate weights, the weights for generating a value for the configuration parameter given a target cardiogram, the generated value indicating the value for the configuration parameter of a source configuration of a target heart from which the target cardiogram is derived; and
one or more processors for executing the computer-executable instructions stored in the one or more computer-readable storage mediums.
11. The one or more computing systems of claim 10 wherein the machine learning training is based on is a neural network.
12. The one or more computing systems of claim 10 wherein the value for the configuration parameter is a classification of the heart.
13. The one or more computing systems of claim 10 wherein the machine learning is based on clustering of the cardiograms.
14. The one or more computing systems of claim 10 wherein at least some of the cardiograms are simulated cardiograms.
15. The one or more computing systems of claim 10 wherein at least some of the cardiograms are collected from patients.
16. The one or more computing systems of claim 10 wherein the computer-executable instructions further control the one or more computing systems to:
access a patient cardiogram of a patient; and
generating a value for the configuration parameter for patient, the value generated based on the weights and the patient cardiogram.
17. A method performed by one or more computing systems for generating a value for a configuration parameter of a heart of a patient, the method comprising:
accessing a patient cardiogram of the patient;
accessing weights that are learned by performing machine learning training, the training based on training cardiograms, each of the training cardiograms derived from electromagnetic output associated with a heart and associated with a value for the configuration parameter of a source configuration of the heart, the electromagnetic output representing electrical activity of the heart; and
generating a value for the configuration parameter of the heart of the patient based on the patient cardiogram and a plurality of the weights, the generated value indicating the value for the configuration parameter of a source configuration of the heart of the patient.
18. The method of claim 17 wherein the machine learning training is based on is a neural network.
19. The method of claim 17 wherein the value for the configuration parameter is a classification.
20. The method of claim 17 wherein the machine learning is based on clustering of the cardiograms.