Naslov
System for Automatic Feature Extraction and Pattern Recognition in EEG Signal Analysis

Autori
Moštak, Ivan ; Friganović, Krešimir ; Zelenika Zeba, Mirta ; Cifrek, Mario

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
7th Croatian Neuroscience Congress - Book of Abstracts / - , 2019, 78-78

Skup
7th Croatian Neuroscience Congress

Mjesto i datum
Zadar, Hrvatska, 12.09.2019. - 15.09.2019

Vrsta sudjelovanja
Poster

Vrsta recenzije
Domaća recenzija

Ključne riječi
Electroencephalography ; Automatic Signal Processing ; Pattern Recognition ; Feature Extraction ; Alpha Spindle

Sažetak
Electrical activity of the brain recorded with the electroencephalogram (electroencephalographic signals, EEG) can be used for extracting features and identifying certain patterns that best describe explicit human psychophysiological states. EEG signals are usually analyzed by neuroscience experts in the fields of medical diagnostics and scientific researchers. Manual analysis of EEG signals is a lengthy process and requires vast expert knowledge. Expert experience and subjective impression can significantly influence the analysis. Different preprocessing steps and the choice of removing different artifacts, like blinking or muscle activity, can include risk of bias. Using MATLAB program package, a system has been developed for automatic EEG signal processing, analysis and feature extraction. EEG analysis consists of automatic loading of signals and accompanying parameters, semi-automatic removal of artifacts (using Independent Component Analysis, ICA) and the process of separating and calculating features. Features that are included in the developed system are: changes in the distribution of characteristic EEG signal bandwidth (Power Spectral Density, PSD), spatio-temporal propagation of brain activity (occipitofrontal direction), special brain waveforms like alpha spindle, determination of individual alpha responsiveness interval and individual alpha peak frequency. Features and repeating patterns that are extracted from the data can be further analyzed in patients with different pathological states (sleep disorders, epilepsy, excessive fatigue, headaches or others). By using automatic signal processing, the analysis is significantly accelerated, and the same criteria for preprocessing and feature extraction is applied to all the EEG signals. Therefore, the likelihood of human error or omission is considerably reduced.

Izvorni jezik
Engleski

Znanstvena područja
Interdisciplinarne prirodne znanosti, Kognitivna znanost (prirodne, tehničke, biomedicina i zdravstvo, društvene i humanističke znanosti)

POVEZANOST RADA