Computer-aided diagnosis of epileptiform transients in EEG

Methods are described for the detection of single-spike as well as spike and wave bursts activity, both of which are abnormal phenomena associated with epilepsy. The algorithms for detecting these transients were developed and tested using real electroencephalogram (EEG) data. The algorithms and their performance are described. The results of the single spike detection test clearly demonstrate the concept of artifact-contaminated EEG; in all tests, at most 17.5% of the detected spikes were real. It is possible, however, that not all detected noise was due to artifact, but that it may have been a result of improper selection of thresholds and parameters prior to detection. This factor may be responsible for missed spikes as well. Spike and wave activity, in contrast to single spikes, contains characteristics that are more discriminable from background activity, among which are its repetitive nature and duration. Because of this it was possible to detect spike and wave activity with very high accuracy.<
>