The goal of this dissertation is to develop an automated multichannel EEG analysis system to assist the EEGer in prescreening epileptic patients. The design concentrates on the problem of detecting epileptic inter-ictal spikes. User-friendly, window-based visualization tools and a multichannel epileptic spike detection system are developed in this study to provide a visualization environment and a spike detection performance better than that of existing systems.; The user-friendly, window-based tools developed here enable the user to simultaneously visualize and manage the results of automated EEG analysis and multichannel EEG data. The new waveform detection algorithm presented here performs the structural analysis of characteristic line segments, obtained through a guided line segment search. This algorithm was implemented and gave results comparable with those obtained from one of the best automated methods in the detection of sleep EEG waveforms. The spike waveform detector using this algorithm made few detections in non-epileptic EEG data. In epileptics, it detected most of epileptic spikes, but it generated false alarms due to epileptic sharp activities and normal EEG activities.; The knowledge-based contextual analysis model, which uses a hypothesis-confirmation process to simulate the EEGer's visual EEG analysis, was used to develop a knowlege-based system for screening out false positive detections generated by the spike waveform detector. The system eliminates most false detections due to normal EEG activities such as alpha, sigma, muscle and eye-movement artifacts. However, it still suffers from false positive detections mainly due to epileptic sharp activities. In two subjects with epilepsy, 72% and 63% of the epileptic spikes visually screened by two EEGers were detected by the system, with 2.68 and 2.89 false detections per minute, respectively. In two subjects with no epileptic spikes agreed by both EEGers, 0.02 and 0.33 false positive detections per minute were obtained.
展开▼
