Development of a Novel Convolution Neural Network for Arrhythmia Classification

Identifying the correct arrhythmia at the time of a clinic event including cardiac arrest is
of high priority to patients, healthcare organizations, and to public health. Recent
developments in artificial intelligence and machine learning are providing new opportunities
to rapidly and accurately diagnose cardiac arrhythmias and for how new mobile health and
cardiac telemetry devices are used in patient care. The current investigation aims to
validate a new artificial intelligence statistical approach called 'convolution neural
network classifier' and its performance to different arrhythmias diagnosed on 12-lead ECGs
and single-lead Holter/event monitoring. These arrhythmias include; atrial fibrillation,
supraventricular tachycardia, AV-block, asystole, ventricular tachycardia and ventricular
fibrillation, and will be benchmarked to the American Heart Association performance criteria
(95% one-sided confidence interval of 67-92% based on arrhythmia type). In order to do so,
the study approach is to create a large ECG database of de-identified raw ECG data, and to
train the neural network on the ECG data in order to improve the diagnostic accuracy.