A tool to perform semi-supervised anomaly detection and annotation on 15 lead ECG signals

Measuring both electrical and mechanical activities of the heart has gained success thanks to technologies able to measure them. Heart electrical activity is measured by means of Electrocardiography which generate Electrocardiographic (ECG) signals. The automatic analysis of ECG signals by means of algorithms and tools may help to detect anomalies and automatically annotate them. Recently, a particular type of network architecture, referred to as Autoencoder (AE), has been used for similar tasks in many fields, both biomedical and not. Nevertheless, using an AE for the analysis of ECG signals can still provide improvements to clinicians. We here present a tool that can be used by clinicians for the semi-automatic identification of anomalous windows in ECG signals. Moreover, the tool allows signal visualization, manual annotation, and measuring.