Abstract: Automatic disease detection is important in relieving the heavy workload of examining prolonged electroencephalograph (EEG). Manually Diagnosing disease in EEG is a tedious process and it consumes tens of hours of EEG recording. Early diagnosis and classification of diseases is very important in clinical practice. With recent development in the biomedical engineering and instruments, EEG recording instruments are able to record the electric activities of brain with high accuracy, which founds EEG as a most important tool for diagnosing the abnormalities of brain. This paper represents automated electroencephalogram based advanced diagnosis of diseases using FastICA (fast independent component analysis) and artificial neural networks (ANNs). FastICA is an efficient method to identify artifact and actual EEG from their mixtures. EEG signals carry the information of human brain with artifacts. These artifacts are removed by FastICA algorithm. Further, an ANN is designed to achieve process like a brain. The clean EEG is fed to feed forward back propagation neural network to diagnosis disease. Training parameters and type of neural networks are decided by operators on the interface. Performance of this model is evaluated using overall accuracy.

Keywords: EEG Signals; Artificial Neural Network; Epilepsy; FastICA; electroencephalograph; ANN; feed forward back propagation.