Classification of event-related potentials in individuals at risk for alcoholism using wavelet transform and artificial neural network

In this study we investigated event related-potential (ERP) component c247, which reflects the recognition process in two classes of subject: a sample with high risk (HR) for alcoholism and a sample of control subjects with low risk (LR). The results of this study suggest that the amplitude of the c247 to repeated pictures of common objects was decreased significantly as compared to the unrepeated pictures for the LR group, but the same was not observed in the HR group. For the matching stimulus we describe an application of an artificial neural network (ANN) technique together with a feature extraction, the wavelet transform, for the classification of ERPs. Two classes of ERP were used: HR for alcoholism and LR for alcoholism. The architecture of the ANN used in the classification is a three-layered feedforward network that implements the backpropagation of error learning algorithm. After training, the network with wavelet coefficients was able to correctly classify over 71% of the HR and LR class of ERP.