DETECTION OF SPIKES WITH ARTIFICIAL NEURAL NETWORKS USING RAW EEG

Artificial neural networks (ANN) using raw electroencephalogram (EEG) data were developed and tested off-line to detect transient epileptifo rm discharges (spike and spike/wave) and EMG activity in an ongoing EE G. In the present study, a feedforward ANN with a variable number of i nput and hidden layer units and two output units was used to optimize the detection system. The ANN system was trained and tested with the b ackpropagation algorithm using a large data set of exemplars. The effe cts of different EEG time windows and the number of hidden layer neuro ns were examined using rigorous statistical tests for optimum detectio n sensitivity and selectivity. The best ANN configuration occurred wit h an input time window of 150 msec (30 input units) and six hidden lay er neurons. This input interval contained information on the wave comp onent of the epileptiform discharge which improved detection. Two-dime nsional receiver operating curves were developed to define the optimum threshold parameters for best detection. Comparison with previous net works using raw EEG showed improvement in both sensitivity and selecti vity. This study showed that raw EEG can be successfully used to train ANNs to detect epileptogenic discharges with a high success rate with out resorting to experimenter-selected parameters which may limit the efficiency of the system. (C) 1998 Academic Press.