Numerical EEG Forward Modeling with Dipolar Sources: H(div) Approach
Miinalainen, Tuuli (2017)
2017
Teknis-luonnontieteellinen
Teknis-luonnontieteellinen tiedekunta - Faculty of Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2017-09-06
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201708241778
https://urn.fi/URN:NBN:fi:tty-201708241778
Tiivistelmä
The objective of this master's thesis project is to study forward electroencephalography (EEG) modeling with divergence conforming, finite element sources. EEG is a method for measuring electric potentials on human head, caused by neural activity in the brain. The main goals were to implement previously studied H(div) - source types to a C++ based toolbox DUNE (Distributed and Unified Numerics Environment), and also to numerically analyze the influence of the element patch size on modeling accuracy. Moreover, an adaptive version of the previously studied H(div) approach is evaluated. The numerical analysis was conducted with a spherical mesh.
The results of the numerical experiments revealed that the divergence conforming source models produce relatively accurate results near the outer gray matter layer boundary. For deeper sources that are located further away from the gray matter boundary, the reference method St. Venant gave more precise results. Moreover, the modeling accuracy for the H(div) source model improved as the size of the element patch grew. Nevertheless, for sources near the gray matter boundary, there were no significant increases in modeling precision detected after taking more than four elements in source configuration. In addition, the adaptive style did not bring any remarkable advantage to the resulting accuracy.
The results of the numerical experiments revealed that the divergence conforming source models produce relatively accurate results near the outer gray matter layer boundary. For deeper sources that are located further away from the gray matter boundary, the reference method St. Venant gave more precise results. Moreover, the modeling accuracy for the H(div) source model improved as the size of the element patch grew. Nevertheless, for sources near the gray matter boundary, there were no significant increases in modeling precision detected after taking more than four elements in source configuration. In addition, the adaptive style did not bring any remarkable advantage to the resulting accuracy.