Analysis of Scale-Invariance in EEG microstates due to Acoustic Stimuli
Author
Lykke, Rasmus Vestergaard
Term
4. semester
Education
Publication year
2022
Pages
102
Abstract
Denne afhandling undersøger, om korte, tilbagevendende mønstre i hjernens aktivitet (EEG-mikrotilstande) ændrer deres selvsimilaritet, når der lyttes til forskellige lyde. For at måle selvsimilaritet blev mikrotilstandssekvenserne omsat til en tilfældig vandring (random walk) og beskrevet med Hurst-eksponenten, som opsummerer langtidssammenhæng i et signal. To estimeringsmetoder blev anvendt: en etableret wavelet-baseret metode (en signalanalysemetode) og en metode med et konvolutionelt neuralt netværk (en type maskinlæringsmodel). Med Hurst-estimater fra begge metoder blev der udført en variansanalyse (ANOVA) på EEG-optagelser fra en lytteopgave for at undersøge, om lydtypen havde en målbar effekt. Resultaterne var ikke entydige, og der er behov for yderligere arbejde.
This thesis examines whether brief, recurring patterns in brain activity (EEG microstates) change their self-similarity when listeners are exposed to different sounds. To quantify self-similarity, the sequences of microstates were mapped to a random walk and summarized by the Hurst exponent, a measure of long-term dependence in a signal. Two estimation approaches were used: an established wavelet-based method (a signal analysis technique) and a method using a convolutional neural network (a type of machine learning model). Using Hurst estimates from both methods, an ANOVA was performed on EEG recordings from a listening task to test whether sound type had a measurable effect. The results were inconclusive, indicating that further work is needed.
[This summary has been rewritten with the help of AI based on the project's original abstract]
Documents