Detecting abnormalities in ECG signals
A method of detecting abnormalities in an ECG signals, comprising receiving an ECG signal, extracting at least one ECG beat the ECG signal; providing the at least one ECG beat to a neural network. Within the neural network, performing at least two layers of convolution operations, and for a beat of the ECG signal, determining a confidence map for the location of each of the P, Q, R, S and T points, determining, from the respective confidence maps the position or absence of points, the points comprising each of the P, Q, R, S and T points; and reporting the positions and/or absence of the points.
1. A method of producing a system for detecting abnormalities in electrocardiogram (ECG) signals, comprising;
providing a neural network configured to:
perform at least two layers of convolution operations;
determine a confidence map for a location of each of a plurality of points;
determine, from the respective confidence map, a position of each of the plurality of points, the points corresponding with one or more of a P, Q, R, S and T feature in each beat of the ECG signal,
providing synthetic training data for the neural network by:
extracting a plurality of single ECG beats from ECG data;
clustering the beats based on their similarity;
sampling the beats uniformly across the clusters;
labelling the sampled beats to provide a set of labelled beats; each labelled beat comprising a position of one or more of a P, Q, R, S, and T features;
producing an expanded set of labelled beats by synthetically modifying the set of labelled beats;
training the neural network using the expanded set of labelled beats, based on an error between an output of the neural network and the positions of the labels in the expanded set of labelled beats.
2. The method of claim 1 , wherein clustering comprises using a k-means algorithm.
3. The method of claim 1 , wherein synthetically modifying the set of labelled beats comprises adding noise.
4. The method of claim 3 , wherein adding noise comprises adding synthetic Gaussian noise.
5. The method of claim 1 , wherein synthetically modifying the set of labelled beats comprises changing the duration of the beat.