EEG allika lokaliseerimine: masinaõppe lähenemisviis
Files
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Inimaju aktiivsuse salvestamise jaoks on olemas mitmeid meetodeid. Üks nendest on EEG, mis suudab ajusignaali mõõta peaaegu samal hetkel, kui see signaal ajus tekib.Samas selle ruumiline täpsus on väga madal. Konkureeriv tehnoloogia on fMRI, mille ruumiline täpsus on hea, kuid ajaline täpsus madal. Mõõtes ajusignaale kasutades mõlemat tehnoloogiat korraga saab kätte signaali, mis on rikas ja täpne aju aktiivsuse kirjeldus nii ruumis kui ka ajas. Signaali allika järeldamist EEG andmetest nimetatakse allika lokaliseerimise probleemiks. Antud uuringus me demonstreerime uut lokaliseerimise meetodit, mis kasutab masinõpet. Uue meetodi suutlikkuse hindamiseks kasutame andmestikku, kus EEG ja fMRI signaalid olid salvestatud samaaegselt. Samuti võrdleme antud töös väljatöötatud meetodit teiste allika lokaliseerimise meetoditega.
There are different techniques for recording human brain activity. One of them EEG can capture brain activity in the time frame at which the activity occurs, but has a poor spatial resolution. Another technology fMRI, captures brain activity with high spatial resolution compared to EEG, but with poor temporal resolution. Simultaneously recording brain activity using these two techniques helps us capture a richer, spatio-temporally more precise description of human brain activity. Inferring the source location within the brain from an EEG signal is defined as EEG source localization problem. In this thesis, a new method that is based on machine learning for solving EEG source localization problem isproposed and its performance is evaluated on a simultaneously recorded EEG and fMRIdata set. This method’s performance is also compared to a commonly used method.
There are different techniques for recording human brain activity. One of them EEG can capture brain activity in the time frame at which the activity occurs, but has a poor spatial resolution. Another technology fMRI, captures brain activity with high spatial resolution compared to EEG, but with poor temporal resolution. Simultaneously recording brain activity using these two techniques helps us capture a richer, spatio-temporally more precise description of human brain activity. Inferring the source location within the brain from an EEG signal is defined as EEG source localization problem. In this thesis, a new method that is based on machine learning for solving EEG source localization problem isproposed and its performance is evaluated on a simultaneously recorded EEG and fMRIdata set. This method’s performance is also compared to a commonly used method.