Detecting abnormalities in ECG signals

A system for improving cardiovascular health, comprising:

• a computer comprising:
• a data reception module for receiving ECG data obtained from a test subject;
• and
• a neural network for determining a final classification comprising a plurality of final classification estimates, each final classification estimate corresponding with a time segment of the ECG signal, wherein the neural network comprises:
• (i) a first subset of layers performing a first series of convolution operations, and determining a preliminary classification comprising a plurality of preliminary classification estimates, each preliminary classification estimate corresponding with a time segment of the ECG signal;
• and
• (ii) a second subset of layers that receive the preliminary classification and an output of a layer of the first subset of layers that precedes a final layer of the first subset of layer, a final layer of the second subset of layers determining a final classification comprising a plurality of final classification estimates, each final classification estimate corresponding with a time segment of the ECG signal,
• wherein at least some of the convolution operations in at least one of the first and the second subset of layers are dilated convolution operations;
• the computer further comprising:
• a reporting module for reporting the final classification estimate to a subject and/or cardiologist.