A retrospective glance at automatic detection of epileptic spike in electroencephalogram

Jing Zhou
2015 European Journal of BioMedical Research  
Automatic detection of epileptic spikes is an important clinical application. It has been developed nearly 40 years. Yet the current automatic detection results are still not as reliable as experienced human interpreters, mainly due to the complex morphology of spikes and the similarity between paroxysmal events in brain activities. By reviewing the previous work, it is noticeable that the implementation of wavelet, ANN and spatiotemporal analysis show promising prospect. By reasonable
more » ... reasonable combination of detection strategies, the detection rate can reach 90% with acceptable false detection rate. Current researches also reveal the reality of lacking of uniform dataset and rules for algorithm comparison. This review aims to compare the pros and cons of current researches and discuss the trend of development in the future. Review far have explored many possibilities of implementing the innovative algorithms in the field of automatic detections. Recently, several researchers have developed sophisticated, staged detection schemes and achieved considerable results. This paper intends to review these attempts made on automatic detection of epileptic spikes, discuss the advantages and disadvantages of the commonly used methodologies, and highlight the progresses. Methods There are two aspects that affect the outcome of a detection process: the feature selection and the classification strategy. Features are usually inherent and distinct properties of the signal, while classification strategy is basically one or several mathematic algorithms aiming to separate data vectors in different classes. Reasonable selection of features is the fundamental of high performance. A sophisticated classification strategy, on the other hand, French Sciences Publishing Group
doi:10.18088/ejbmr.1.4.2015.pp9-17 fatcat:s6f2nwblcjhvldxd7kjzfput3e