The proposed research will investigate the development of an """"""""intelligent"""""""", multichannel (16 channels) automated (general purpose computer-based) system for the reliable detection of epileptogenic sharp waveforms (spikes and shart waves-STs) in the electroencephalogram (EEG). The system will provide the time of occurrence of individual STs, their spatial location, along with values of pertinent electrographic characteristics (e.g. amplitude, duration) thereof. Special attention will be given to the minimization of false alarms due to artifact and normal EEG activity. Such a system will especially facilitate the rapid and reliable analysis of long-term EEG recordings from subjects with known or suspected epilepsy, thus contributing to their optimum therapeutic management as well as resulting in savings of time and money. Extensive heuristics based on human pattern recognition, which involves contextbased analysis, will be utilized. In this way, the system will incorporate the intelligence of an electroencephalographer (EEGer). An Artificial Intelligence approach will be implemented, whereby EEG features quantifying candidate waveforms, as well as automated assessment of the subject's state of consciousness will be presented to an Expert System for eventual decision making. The knowledge base of the Expert system will be developed in close cooperation with an experienced EEGer, so that the extensive contextbased heuristics utilized in the visual analysis process (e.g. spatio-temporal processing involving multichannel correlations) will be implemented. Special emphasis will be given to programming into the detection system the ability to discriminate against ST-like artifactual activity in the EEG record (e.g. electrocardiographic and electromyographic interference) as well as against sharp but normal EEG activity (e.g. sleep spindles). Extensive validation studies will be conducted utilizing EEG data from epileptic subjects, other patients, and normal controls. An investigation will also take place of the possibility of implementing the detection system in a microprocessor-based version for portability and ease of operation.
| Jansen, B H; Dawant, B M (1989) Knowledge-based approach to sleep EEG analysis--a feasibility study. IEEE Trans Biomed Eng 36:510-8 |
| Glover Jr, J R; Raghavan, N; Ktonas, P Y et al. (1989) Context-based automated detection of epileptogenic sharp transients in the EEG: elimination of false positives. IEEE Trans Biomed Eng 36:519-27 |
