Neural Network Detector of ECG Signal Distortions

Abstract

The use of artificial neural networks for detection of ECG signal distortions was discussed. Training and test databases were compiled. A technique for analysis of training samples based on the k-means clustering method was suggested. The effect of the number of hidden layer neurons on the neural network efficiency was studied. A method for testing the neural network efficiency based on the receiver operating characteristic (ROC) curve was developed. The structural principle of the neural network detector of ECG signal distortions was also developed. Testing of the system demonstrated high values of sensitivity and specificity (94.5%), as well as a high mean value of AUC (0.97).