PT  - JOURNAL ARTICLE
AU  - M. A. Pinto-Orellana
AU  - F. R. Cerqueira
TI  - Filter-banks and artificial intelligence in seizure detection using electroencephalograms
AID  - 10.1101/105650
DP  - 2017 Jan 01
TA  - bioRxiv
PG  - 105650
4099  - http://biorxiv.org/content/early/2017/02/03/105650.short
4100  - http://biorxiv.org/content/early/2017/02/03/105650.full
AB  - Epilepsy is the most typical neurological disease in the world, and it implies an expensive and specialized diagnosis process based on electroencephalograms and video recordings. We developed a method that only requires the brainwave provided by the difference between two standard-located electrodes. Our proposed technique separates the original signal using a filter array with three different types of filters, and then extracts several features based on information theory and statistical information. In our study, we found that only 10 characteristics, of which the most important are related to higher frequencies, are required to offer an accuracy of 94%, a specificity of 95% and a sensitivity of 87% using C4.5 decision trees.