Linda K. Dance

Predictability analysis for an automated seizure prediction algorithm.

Epileptic seizures of mesial temporal origin are preceded by changes in signal properties detectable in the intracranial EEG. A series of computer algorithms designed to detect the changes in spatiotemporal dynamics of the EEG signals and to warn of impending seizures have been developed. In this study, we evaluated the performance of a novel adaptive threshold seizure warning algorithm (ATSWA), which detects the convergence in Short-Term Maximum Lyapunov Exponent (STLmax) values among critical intracranial EEG electrode sites, as a function of different seizure warning horizons (SWHs). The ATSWA algorithm was compared to two statistical based naïve prediction algorithms (periodic and random) that do not employ EEG information. For comparison purposes, three performance indices “area above ROC curve” (AAC), “predictability power” (PP) and “fraction of time under false warnings” (FTF) were defined and the effect of SWHs on these indices was evaluated. The results demonstrate that this EEG based seizure warning method performed significantly better (P < 0.05) than both naïve prediction schemes. Our results also show that the performance indexes are dependent on the length of the SWH. These results suggest that the EEG based analysis has the potential to be a useful tool for seizure warning.

Data Mining in EEG: Application to Epileptic Brain Disorders

Epilepsy is one of the most common brain disorders. At least 40 million people or about 1% of the population worldwide currently suffer from epilepsy Despite advances in neurology and neuroscience, approximately 25–30% of epileptic patients remain unresponsive to anti-epileptic drug treatment, which is the standard therapy for epilepsy There is a growing body of evidence and interest in predicting epileptic seizures using intracranial electroencephalogram (EEG), which is a tool for evaluating the physiological state of the brain. Although recent studies in the EEG dynamics have been used to demonstrate seizure predictability, the question of whether the brain’s normal and pre-seizure epileptic activities are distinctive or differentiable remains unanswered. In this study, we apply data mining techniques to EEG data in order to verify the classifiability of the brain dynamics. We herein propose a quantitative analysis derived from the chaos theory to investigate the brain dynamics. We employ measures of chaoticity and complexity of EEG signals, including Short-Term Maximum Lyapunov Exponents, Angular Frequency, and Entropy, which were previously shown capable of contemplating dynamical mechanisms of the brain network. Each of these measures can be used to display the state transition toward seizures, in which different states of patients can be classified (normal, pre-seizure, and post-seizure states). In addition, optimization and data mining techniques are herein proposed for the extraction of classifiable features of the brain’s normal and pre-seizure epileptic states from spontaneous EEG. We use these features in study of classification of the brain’s normal and epileptic activities. A statistical cross validation is implemented to estimate the accuracy of the brain state classification. The results of this study indicate that it may be possible to design and develop efficient seizure warning algorithms for diagnostic and therapeutic purposes.

Methodology for the 3D modeling and visualization of concurrency networks

One of the primary formalisms for modeling concurrency and resource contention in systems is Petri nets. The literature on Petri nets is rich with activity on applications as well as extensions. We use the basic Petri net formalism as well as several extensions to demonstrate how metaphor can be applied to yield 3D model worlds. A number of metaphors, including 3D-primitive and landscape are employed within the framework of VRML-enabled simulation. We designed a template for use in creating any Petri net model and then using the template, implemented an example model utilizing metaphors for both the structure and the behaviors of the model. We determined that the result is an effectively and efficiently communicated model with high memory retention. The modeling methodology that we employ was successfully implemented for Petri nets with the ability for model reuse and/or personalization with any aesthetics applied to the desired Petri net.

Automated Seizure Prediction Algorithm and its Statistical Assessment: A Report from Ten Patients

The ability to predict epileptic seizures well prior to their clinical onset provides promise for new diagnostic applications and novel approaches to seizure control. Several groups of investigators have reported that it may be possible to predict seizures based on the quantitative analysis of EEG signal characteristics. The objective of this chapter is first to report an automated seizure warning algorithm, and second to compare its performance with other, theoretically sound, statistical algorithms. The proposed automated seizure prediction algorithm (ASPA) consists of an optimization method for the selection of critical cortical sites using measures from nonlinear dynamics, and a novel method for the detection of preictal transitions using adaptive transition thresholds according to the current state of dynamical interactions among brain sites. Continuous long-term (mean 210 hours per patient) intracranial EEG recordings obtained from ten patients with intractable epilepsy (total of 130 recorded seizures) were analyzed to test the proposed algorithm. For each patient, the prediction ROC (receiver operating characteristic) curve, generated from ASPA, was compared with the ones from periodic and random prediction schemes. The results showed that the performance of ASPA is significantly superior to each naive prediction method used (p-value < 0.05). This suggests that the proposed nonlinear dynamical analysis of EEG contains relevant information to prospectively predict an impending seizure, and thus has potential to be useful in clinical applications.