Date of Award

January 2013

Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical Engineering

First Advisor

Reza Fazel-Rezai


This dissertation presents innovative approaches based on fuzzy logic in epileptic seizure detection and prediction from Electroencephalogram (EEG). The fuzzy rule-based algorithms were developed with the aim to improve quality of life of epilepsy patients by utilizing intelligent methods. An adaptive fuzzy logic system was developed to detect seizure onset in a patient specific way. Fuzzy if-then rules were developed to mimic the human reasoning and taking advantage of the combination in spatial-temporal domain. Fuzzy c-means clustering technique was utilized for optimizing the membership functions for varying patterns in the feature domain. In addition, application of the adaptive neuro-fuzzy inference system (ANFIS) is presented for efficient classification of several commonly arising artifacts from EEG. Finally, we present a neuro-fuzzy approach of seizure prediction by applying the ANFIS. Patient specific ANFIS classifier was constructed to forecast a seizure followed by postprocessing methods. Three nonlinear seizure predictive features were used to characterize changes prior to seizure. The nonlinear features used in this study were similarity index, phase synchronization, and nonlinear interdependence. The ANFIS classifier was constructed based on these features as inputs. Fuzzy if-then rules were generated by the ANFIS classifier using the complex relationship of feature space provided during training. In this dissertation, the application of the neuro-fuzzy algorithms in epilepsy diagnosis and treatment was demonstrated by applying the methods on different datasets. Several performance measures such as detection delay, sensitivity and specificity were calculated and compared with results reported in literature. The proposed algorithms have potentials to be used in diagnostics and therapeutic applications as they can be implemented in an implantable medical device to detect a seizure, forecast a seizure, and initiate neurostimulation therapy for the purpose of seizure prevention or abortion.