To search, Click below search items.


All Published Papers Search Service


Hybrid Two Stage Neuro Genetic System for Arrhythmia Diagnosis


Hela Lassoued and Raouf Ketata


Vol. 18  No. 9  pp. 31-42


This work plans to design an intelligent Electrocardiogram (ECG) diagnosis support system that can identify heart abnormalities with high accuracy (ACC), low normalized mean square error (NMSE), and fast classification time response (TR). Therefore, 12 proposed Multilayer Perceptron networks (MLP) architectures, based on Single Stage MLP and Two Stage MLP, were evaluated and compared for their ability to classify five cardiac classes. The training and testing ECG signals were obtained from MIT-BIH database. The network inputs are either the entire feature dataset, containing 62 features based on morphological and Discrete Wavelet Transformer (DWT) coefficients, or a selected feature dataset, which is acquired after applying the Genetic Algorithm (GA). Concerning the number of hidden neurons, the test and error method and the GA designed for the network optimization were evaluated. The obtained performances were compared and discussed. Among the different networks, the proposed two-stage MLP network, called Net6, which uses the selected ECG features and applies the GA to fix its hidden neurons number, possesses the highest ACC, a lower NMSE and an acceptable classification time. Therefore, this network proves to be a suitable classifier in ECG diagnosis system especially when the analysis requires an accurate result and no matter of the classification time response.


Arrhythmia diagnosis, hybrid, neuro, genetic, Two Stage MLP.