A system for detecting epileptiform discharges in the EEG: real-time operation and clinical trial

The authors have developed a real-time PC-based multistage system for automated detection of epileptiform discharges (EDs)-spikes and sharpwaves-in routine and long-term EEG recordings. A central component is an expert system which uses spatial and temporal contextual cues in the EEG to detect EDs and to reject non-epileptiform waves. The spike detection system has been evaluated in a major clinical study by comparison of automated analysis with independent results from three electroencephalographers (EEGers). The data comprised 521 consecutive routine EEG recordings (173 hours). The system detected 36 of the 38 EEGs containing definite EDs-a sensitivity of 95%-had an average false detection rate of 0.29 EDs/hour, less than 1% of that reported in studies of other ED detectors. The results of this study confirm that the authors´ system is at the forefront of efforts to automate the detection of epileptiform activity in the EEG