Multi-channel and with rhythm transfer learning

Abstract

Techniques for classifying cardiac events in electrocardiogram (ECG) data. A feature set is generated by analyzing ECG data for a patient using a first phase in a machine learning architecture. A first cardiac event in the ECG data is classified based on the feature set, using the first phase in the machine learning architecture. A second cardiac event in the ECG data is classified based on the classified first cardiac event and the feature set, using a second phase in the machine learning architecture. The second cardiac event overlaps at least partially in time with the first cardiac event. Further, a plurality of feature sets, corresponding to a plurality channels of ECG data, are generated using paths in a machine learning architecture. A cardiac event in the ECG data is classified using the machine learning architecture based on the plurality of feature sets.

We claim: 
     
         1 . A computer-implemented method, comprising:
 generating a feature set by analyzing electrocardiogram (ECG) data for a patient using a first phase in a machine learning architecture, the first phase comprising a deep learning neural network;   classifying, based on the feature set and using the first phase, a first cardiac event in the ECG data, wherein the first phase is configured to classify cardiac events of a plurality of types;   classifying, based on the feature set and the classified first event and using a second phase in the machine learning architecture, a second cardiac event in the ECG data, wherein the second phase is configured to classify cardiac events of a single type; and   classifying, based on the feature set and the classified first event and using a third phase in the machine learning architecture, a third cardiac event in the ECG data, wherein the third phase is configured to classify cardiac events of a single type different from the second phase, and wherein the second and third cardiac events occur in the patient at least partially at the same time as the first cardiac event.   
     
     
         2 . The computer-implemented method of  claim 1 , wherein the classified first event the classified second event, and the classified third event facilitate medical treatment of the patient. 
     
     
         3 . The computer-implemented method of  claim 1 , further comprising:
 receiving the ECG data for the patient at a server, wherein the ECG data are collected from the patient using a sensor device, transmitted from the sensor device to a mobile device, and transmitted to the server using the mobile device.   
     
     
         4 . The computer-implemented method of  claim 1 , wherein the second phase is configured to identify a first pre-determined single type of cardiac rhythm and wherein the third phase is configured to identify a second pre-determined single type of cardiac rhythm, different from the first type. 
     
     
         5 . The computer-implemented method of  claim 4 , wherein the second and third phases are configured to each provide a binary determination as to whether the ECG data comprises the respective pre-determined type of cardiac rhythm. 
     
     
         6 . The computer-implemented method of  claim 1 , wherein the machine learning architecture comprises a plurality of additional phases, and wherein each additional phase is configured to identify a different pre-determined single type of cardiac event occurring simultaneously with the first cardiac event. 
     
     
         7 . The computer-implemented method of  claim 1 , wherein the classifying the second cardiac event further comprises:
 providing the classified first cardiac event and the feature set as input to a same layer in the second phase.   
     
     
         8 . The computer-implemented method of  claim 7 , wherein the layer comprises at least one of a fully connected layer or a softmax layer. 
     
     
         9 . The computer-implemented method of  claim 1 , wherein the deep learning neural network comprises a convolutional neural network comprising a set of connected layers comprising:
 a convolution layer;   a batch normalization layer;   an activation function layer; and   a regularization layer.   
     
     
         10 . The computer-implemented method of  claim 9 , wherein the regularization layer comprises a dropout layer. 
     
     
         11 . The computer-implemented method of  claim 1  wherein the machine learning architecture comprises a supervised machine learning architecture, and wherein the second phase and the third phase are each trained using a trained first phase of the machine learning architecture. 
     
     
         12 . A non-transitory computer-readable medium containing computer program code that, when executed by operation of a computer processor, performs an operation comprising:
 generating a feature set by analyzing electrocardiogram (ECG) data for a patient using a first phase in a machine learning architecture, the first phase comprising a deep learning neural network;   classifying, based on the feature set and using the first phase, a first cardiac event in the ECG data, wherein the first phase is configured to classify cardiac events of a plurality of types;   classifying, based on the feature set and the classified first event and using a second phase in the machine learning architecture, a second cardiac event in the ECG data, wherein the second phase is configured to classify cardiac events of a single type; and   classifying, based on the feature set and the classified first event and using a third phase in the machine learning architecture, a third cardiac event in the ECG data,   wherein the third phase is configured to classify cardiac events of a single type different from the second phase, and   wherein the second and third cardiac events occur in the patient at least partially at the same time as the first cardiac event.   
     
     
         13 . The non-transitory computer-readable medium of  claim 12 , wherein the classified first event the classified second event, and the classified third event facilitate medical treatment of the patient. 
     
     
         14 . The non-transitory computer-readable medium of  claim 12 ,
 wherein the second phase is configured to provide a binary determination as to whether the ECG data comprises a first pre-determined single type of cardiac rhythm, and   wherein the third phase is configured to provide a binary determination as to whether the ECG data comprises a second pre-determined single type of cardiac rhythm, different from the first type.   
     
     
         15 . The non-transitory computer-readable medium of  claim 12 , wherein the classifying the second cardiac event further comprises:
 providing the classified first cardiac event and the feature set as input to a same layer in the second phase.   
     
     
         16 . A system, comprising:
 a computer processor; and   a memory having instructions stored thereon which, when executed on the computer processor, performs an operation comprising:
 generating a feature set by analyzing electrocardiogram (ECG) data for a patient using a first phase in a machine learning architecture, the first phase comprising a deep learning neural network; 
 classifying, based on the feature set and using the first phase, a first cardiac event in the ECG data, wherein the first phase is configured to classify cardiac events of a plurality of types; 
 classifying, based on the feature set and the classified first event and using a second phase in the machine learning architecture, a second cardiac event in the ECG data, wherein the second phase is configured to classify cardiac events of a single type; and 
 classifying, based on the feature set and the classified first event and using a third phase in the machine learning architecture, a third cardiac event in the ECG data,
 wherein the third phase is configured to classify cardiac events of a single type different from the second phase, and 
 wherein the second and third cardiac events occur in the patient at least partially at the same time as the first cardiac event. 
 
   
     
     
         17 . The system of  claim 16 , wherein the classified first event the classified second event, and the classified third event facilitate medical treatment of the patient. 
     
     
         18 . The system of  claim 16 ,
 wherein the second phase is configured to provide a binary determination as to whether the ECG data comprises a first pre-determined single type of cardiac rhythm, and   wherein the third phase is configured to provide a binary determination as to whether the ECG data comprises a second pre-determined single type of cardiac rhythm, different from the first type.   
     
     
         19 . The system of  claim 16 , wherein the classifying the second cardiac event further comprises:
 providing the classified first cardiac event and the feature set as input to a same layer in the second phase.   
     
     
         20 . The system of  claim 16 , the operation further comprising:
 receiving the ECG data for the patient at a server, wherein the ECG data are collected from the patient using a sensor device, transmitted from the sensor device to a mobile device, and transmitted to the server using the mobile device.