Automated ventricular ectopic beat classification

1. A computer-implemented method for classifying heartbeats using patient electrocardiogram (ECG) data, comprising:

receiving the ECG data, the ECG data comprising unstructured waveform data and corresponding structured time interval data relating to a plurality of heartbeats, wherein a mobile device is configured to receive the ECG data from a remote sensor device;

providing the unstructured waveform data to a convolutional neural network in a first path of a machine learning architecture;

generating a first plurality of output values by analyzing the unstructured waveform data, without the corresponding structured time interval data, using the convolutional neural network in the first path of the machine learning architecture;

providing the structured time interval data to a fully-connected neural network in a second path of the machine learning architecture;

generating a second plurality of output values by analyzing the structured time interval data, without the corresponding unstructured waveform data, using the fully-connected neural network in the second path of the machine learning architecture; and

classifying one or more of the plurality of heartbeats in the ECG data by concatenating the first plurality of output values generated by the convolutional neural network using the waveform in data and the second plurality of output values generated by the fully connected neural network using the corresponding time interval data, wherein the convolutional neural network comprises a different structure from the fully connected neural network.

2. The computer-implemented method of claim 1 , wherein the first plurality of output values and the second plurality of output values are concatenated using a fully connected layer in the machine learning architecture.

3. The computer-implemented method of claim 1 , wherein the convolutional neural network comprises a set of connected layers comprising:

a convolution layer;

a batch normalization layer;

an activation function layer; and

a regularization layer.

4. The computer implemented method of claim 3 , wherein the set of connected layers is repeated in the convolutional neural network.

5. The computer implemented method of claim 4 , wherein the set of connected layers is repeated at least ten times in the convolutional neural network.

6. The computer implemented method of claim 4 , wherein the convolutional neural network further comprises a plurality of convolution layers, a plurality of batch normalization layers, and a plurality of activation function layers, in addition to the set of connected layers.

7. The computer-implemented method of claim 1 , wherein the second path does not comprise a convolution layer and wherein the second path comprises a deep fully-connected neural network.

8. The computer-implemented method of claim 1 , wherein classifying one or more of the plurality of heartbeats in the ECG data comprises: verifying, using the machine learning architecture, that the heartbeat comprises a beat, based on the concatenated first plurality of output values and second plurality of output values.

9. The computer-implemented method of claim 1 , further comprising:

processing the ECG data to generate the waveform data and the time interval data, the processing comprising identifying the plurality of heartbeats in the ECG data.

10. The computer-implemented method of claim 1 , wherein classifying one or more of the plurality of heartbeats in the ECG data comprises classifying the one or more heartbeats as Ventricular Ectopic Beats (VEBs).

11. A computer program product for classifying heartbeats using patient electrocardiogram (ECG) data, the computer program product comprising:

a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to perform an operation, the operation comprising:

receiving the ECG data, the ECG data comprising unstructured waveform data and corresponding structured time interval data relating to a plurality of heartbeats;

providing the unstructured waveform data to a convolutional neural network in a first path of a machine learning architecture based on the convolutional neural network being configured to analyze unstructured data;

generating a first plurality of output values by analyzing the unstructured waveform data, without the corresponding structured time interval data, using the convolutional neural network in the first path of the machine learning architecture;

providing the structured time interval data to a fully-connected neural network in a second path of the machine learning architecture based on the fully-connected neural network being configured to analyze structured data;

generating a second plurality of output values by analyzing the structured time interval data, without the corresponding unstructured waveform data, using the fully-connected neural network in the second path of the machine learning architecture; and

classifying a heartbeat of the plurality of heartbeats in the ECG data by concatenating the first plurality of output values generated by the convolutional neural network using the waveform data and the second plurality of output values generated by the fully-connected neural network using the corresponding time interval data, using the machine learning architecture, wherein the convolutional neural network comprises a different structure from the fully connected neural, network, comprising:

verifying, using the machine learning architecture, that the heartbeat comprises a beat, based on the concatenated first plurality of output values and second plurality of output values; and

determining, based on the concatenated first plurality of output values and second plurality of output values, a probability that the heartbeat comprises a ventricular ectopic beat (VEB).

12. The computer program product of claim 11 , wherein the first plurality of output values and the second plurality of output values are concatenated using a fully connected layer in the machine learning architecture.

13. The computer program product of claim 11 , wherein the convolutional neural network comprises a set of connected layers comprising:

a convolution layer;

a batch normalization layer;

an activation function layer; and

a regularization layer, and

wherein the set of connected layers is repeated in the convolutional neural network.

14. The computer program product of claim 11 , wherein classifying one or more of the plurality of heartbeats in the ECG data comprises identifying a cardiac irregularity based on the ECG data, the operation further comprising:

providing treatment to a patient for the cardiac irregularity based on the classified one or more heartbeats.

15. A system, comprising:

a processor; and

a memory storing a program, which, when executed on the processor, performs an operation, the operation comprising:

receiving patient electrocardiogram (ECG) data;

pre-processing the ECG data, comprising:

detecting a plurality of heartbeats based on the ECG data;

generating unstructured waveform data relating to the detected plurality of heartbeats; and

generating corresponding structured time interval data relating to the plurality of heartbeats;

providing the unstructured waveform data to a convolutional neural network in a first path of a machine learning architecture based on the convolutional neural network being configured to analyze unstructured data;

generating a first plurality of output values by analyzing the unstructured waveform data, without the corresponding structured time interval data, using the convolutional neural network in the first path of the machine learning architecture;

providing the structured time interval data to a fully-connected neural network in a second path of the machine learning architecture based on the fully-connected neural network being configured to analyze structured data;

generating a second plurality of output values by analyzing the structured time interval data, without the corresponding unstructured waveform data, using the fully-connected neural network in the second path of the machine learning architecture; and

classifying a heartbeat of the plurality of heartbeats in the ECG data by concatenating the first plurality of output values generated by the convolutional neural network using the waveform data and the second plurality of output values generated by the fully-connected neural network using the corresponding time interval data, wherein the convolutional neural network comprises a different structure from the fully connected neural network.

16. The system of claim 15 , wherein the first plurality of output values and the second plurality of output values are concatenated using a fully connected layer in the machine learning architecture.

17. The system of claim 15 , wherein the convolutional neural network comprises a set of connected layers comprising:

a convolution layer;

a batch normalization layer;

an activation function layer; and

a regularization layer, and

wherein the set of connected layers is repeated in the convolutional neural network.

18. The system of claim 17 , wherein the set of connected layers is repeated at least ten times in the convolutional neural network.

19. The system of claim 15 , wherein classifying one or more of the plurality of heartbeats in the ECG data comprises verifying, using the machine learning architecture, that the heartbeat comprises a beat, based on the concatenated first plurality of output values and second plurality of output values.

20. The system of claim 15 , wherein classifying one or more of the plurality of heartbeats in the ECG data comprises classifying the one or more heartbeats as Ventricular Ectopic Beats (VEBs).