Atrial Arrhythmias detection based on neural network combining fuzzy classifiers

Accurate detection of atrial arrhythmias is important for implantable devices to treat them. A novel method is proposed to identify sinus rhythm, atrial flutter and atrial fibrillation. Here three different feature sets are firstly extracted based on the frequency-domain, the time-frequency domain and the symbolic dynamics. Then a classifier with two sub-layers is proposed. Three fuzzy classifiers are used as the first layer to perform pre-classification task corresponding to different feature sets respectively. A multilayer perceptron neural network is used as the final classifier. The performance of this algorithm is evaluated with two databases. One is the MIT-BIH arrhythmia database and the other is the endocardial electrogram database. A comparative assessment of the performance of the proposed classifier with individual fuzzy classifier shows that the algorithm can improve the overall accuracy for atrial arrhythmias classification. The implementation of this algorithm in implantable devices may provide accurate detection of atria] affhythmias.