Algorithm for ECG forms clustering based on artificial neural networks

The possibilities of machine learning methods and artificial neural networks use to solve problems of blind ECG cardiocycles clustering were explored. An algorithm for ECG cardiocycles clustering is based on the Kohonen neural network and k-means method is developed. The proposed algorithm is evaluated on annotated standard database MGH/MF waveform database. The accuracy indicators of this algorithm is Acc − 96,92%. This algorithm can be used as the basis for ECG diagnostic software.