System and methods for electrocardiogram beat similarity analysis using deep neural networks

A method comprising:

• selecting a first beat and a second beat from a plurality of beats;
• determining a dynamic time warping (DTW) distance between the first beat and the second beat;
• setting a similarity label for the first beat and the second beat based on the DTW distance;
• mapping the first beat to a first feature space embedding using the deep neural network;
• mapping the second beat to a second feature space embedding using the deep neural network;
• calculating a loss using the first feature space embedding, the second feature space embedding, and the similarity label;
• and
• adjusting weights and biases of the deep neural network based on the loss.

An electrocardiogram (ECG) processing device comprising:

• a display device;
• a user input device;
• a memory storing a deep neural network and instructions;
• and
• a processor communicably coupled to the display device, the user input device, and the memory, and when executing the instructions, configured to:
• select a first beat and a second beat from a plurality of ECG beat data;
• determine a dynamic time warping (DTW) distance between the first beat and the second beat;
• set a similarity label for the first beat and the second beat based on the DTW distance and a DTW distance threshold, wherein the DTW distance threshold is set based on user input received from the user input device;
• train the deep neural network using at least the first beat, the second beat, and the similarity label, to produce a trained deep neural network;
• receive Holter monitor data;
• separate the Holter monitor data into a plurality of beats;
• map the plurality of beats to a plurality of feature space embeddings using the trained deep neural network;
• cluster the plurality of beats based on the plurality of feature space embeddings to produce a plurality of clusters;
• assign physiological labels to the plurality of clusters;
• and
• display a plurality of representative beats corresponding to the plurality of clusters, along with corresponding physiological labels via the display device.