| EEG-based epilepsy diagnosis and seizure detection is still in its early experimental stages. In this research, a multi-paradigm approach is advocated, integrating three novel computational paradigms: wavelet transforms, chaos theory, and artificial neural networks.; First, a novel wavelet-chaos methodology is presented for analysis of EEGs and delta, theta, alpha, beta, and gamma sub-bands of EEGs for detection of seizure and epilepsy. The methodology is applied to three different groups of EEG signals: (a) healthy subjects, (b) epileptic subjects during a seizure-free interval (interictal EEG), and (c) epileptic subjects during a seizure (ictal EEG). The effectiveness of the correlation dimension (CD) and largest Lyapunov exponent (LLE) in differentiating between the three groups is investigated based on statistical significance of the differences. This research challenges the assumption that the EEG represents the dynamics of the entire brain as a unified system and needs to be treated as a whole. It is indeed found that the sub-bands yield more accurate information about constituent neuronal activities underlying the EEG and consequently, certain changes in the EEGs that are not evident in the original full-spectrum EEG are amplified when each sub-band was analyzed separately. Moreover, it is concluded that for the higher frequency beta and gamma sub-bands, the CD differentiates between the three groups, whereas for the lower frequency alpha sub-band, the LLE differentiates between the three groups.; Based on potential markers of abnormality discovered using the wavelet-chaos methodology, a novel wavelet-chaos-neural network methodology is developed for classification of EEGs into healthy, ictal, and interictal EEGs. Three parameters are employed for EEG representation: standard deviation (STD, quantifying the signal variance), CD, and LLE (quantifying the non-linear chaotic dynamics of the signal). Based on extensive parameteric analysis, it is concluded that all three key components of the wavelet-chaos-neural network methodology are important for improving the EEG classification accuracy. It was discovered that a particular mixed-band feature space consisting of nine parameters and the Levenberg-Marquardt Backpropagation Neural Network (LMBPNN) result in the highest classification accuracy, a high value of 96.7%.; To increase the robustness of classification, a novel principal component analysis (PCA)-enhanced cosine radial basis function neural network classifier is developed. For epilepsy diagnosis, when only normal and interictal EEGs are considered, the classification accuracy of the proposed model was 99.3%. This statistic is especially remarkable because even the most highly trained neurologists do not appear to be able to detect interictal EEGs more than 80% of the time.; Next, an efficient spiking neural network (SNN) model is presented for epilepsy and epileptic seizure detection using three training algorithms: SpikeProp (using both incremental and batch processing), QuickProp, and RProp. It is concluded that RProp is the best training algorithm because it has the highest classification accuracy among all training algorithms especially for large size training datasets with about the same computational efficiency provided by SpikeProp. The SNN model for EEG classification and epilepsy and seizure detection employs RProp as training algorithm and yields a classification accuracy of 92.5%.; Finally, to reach the next milestone in artificial intelligence, a biologically realistic neural network, a new Multi-Spiking Neural Network (MuSpiNN) model is presented where information from one neuron is transmitted to the next in the form of multiple spikes via multiple synapses. (Abstract shortened by UMI.)... |
