Abstracts

Rationale: Timely detection of occurrence of seizures in epileptic patients and alarming caregivers for prompt intervention is extremely important in epilepsy care. It would drastically improve the quality of life of epileptic patients and reduce mortality. Despite remarkable past and current research and development in this field, there are no, to our knowledge, algorithms that could reliably and promptly detect the occurrences of epileptic seizures by real-time EEG analysis. They have limited ability to extract full information contained in activity patterns of highly irregular and non-stationary signals produced by biological sources. Besides, existing methods in general require quite long EEG fragments, or epochs, to be processed - sometimes tens of seconds. This makes these methods hardly acceptable when the detection of seizure must be as quick as possible, like in seizure alarm systems. Methods: We propose a novel method for real-time electrographic seizure detection. The key part of the method is decomposition of the EEG signal into elementary components (Fragmentary decomposition, FD), using original technique of short-term Fourier transform - the Similar basis functions algorithm (Melkonian 2010). FD creates remarkably accurate explicit model of the signal under analysis and provides opportunity for computer reconstructions of different sets of defined signal components. This approach provides a more elaborate way for waveform analysis which identifies specific shape of each peak in the time course of non-stationary signal. The components of the model signal are then processed by a specific pattern recognition algorithm, which may be tuned for recognition of any specific combination of model components with desired accuracy. Results: The method was tested on EEG recordings from genetically epileptic rats of WAG/Rij strain (the animal model of absence epilepsy, 8 animals), and human EEG recordings from patients with absence epilepsy (4 recordings). The rat EEGs contained from tens to hundreds of spike-wave discharges (SWD), while human recordings contained 2-4 SWDs. The EEG recordings were ‘fed' to the detection algorithm with the same sampling rate as was used during acquisition to simulate the real-time processing. In all cases the method robustly detected 100% of SWDs, with no false-positive detections in human and few ones in rat EEGs (mostly during very strong sleep spindles). Besides ‘normal' long SWDs, the short and relatively weak SWDs (containing two and more spike-waves) were also successfully detected. The detection time was about 200-500 ms from the seizure onset in rats, and about 700-1000 ms in humans (this difference is due to the difference in SWD frequency, which is higher in rats). Conclusions: The proposed method may be used for fast and reliable detection of electrographic seizures in the EEG. Its ability to detect short epileptogenic events and other specific patterns (which may be small in amplitude) makes the method useful in research directed to seizure prediction.