Anomaly Detection in Cardiac Related Datasets

Abstract

cardiovascular disease is one of the most common diseases in the modern world. If recognized early, then it can significantly reduce the damage to the patient. This work describes the detection of anomalies in electrocardiogram (ECG) readings. Anomaly detection in data mining finds instances, occurrences, and observations that differ from a dataset's regular pattern of activity. Using the ECG dataset as input, the initial step in this method is signal pre-processing. High pass, low pass, and notch filters are used to de-noise ECG signals as part of the signal pre-processing. ECG signal de-noising is a significant pre-processing step that highlights the characteristic waves in ECG data while attenuating the disturbances. The emergence of the ECG signal coefficients from signal pre-processing is trained and tested in the second stage. The classification of the ECG signal using LSTM RNN Model is the final stage. Recurrent neural networks are artificial neural networks that use sequential data or time series data (RNN). The LSTM RNN Model effectively separates out extraneous data, prevents signal information loss, lowers computational complexity, and classifies the ECG signal.