Inventing the future of neurology: Integrated wavelet-chaos-neural network models for knowledge discovery and automated EEG-based diagnosis of neurological disorders

The author has been advancing a multi-paradigm integrated approach for solution of complicated and intractable dynamic pattern recognition problems. The focus of this keynote lecture is data mining and knowledge discovery from time-series signals obtained from complex phenomena. Novel wavelet-chaos-neural network models are presented for signal processing of brain waves as recorded by electroencephalographs (EEGs) for automated EEG-based diagnosis of neurological disorders such as epilepsy and the Alzheimer’s disease (AD). Through extensive parametric studies and information reuse and integration certain combinations of parameters from the EEG sub-bands were discovered to be effective markers for seizure detection and epilepsy diagnosis. The model can distinguish among healthy, interictal, and ictal EEGs with a high accuracy of more than 96% substantially better than practicing neurologists and epileptologists. The extension the methodology for early onset diagnosis of the AD will be delineated.