Low-Dimensional Dynamics Encoding in Human Brain Data Open Access

Yating, Yang (Spring 2018)

Permanent URL: https://etd.library.emory.edu/concern/etds/df65v784d?locale=en


The brain is a complex, multiscale structure where different levels work together to produce behaviors and cognition. People have placed much emphasis on the functioning of individual parts of the brain, but in order to achieve one of the ultimate goals of neuroscience, we need to consider the brain dynamics as a whole, and focus not only on each part of the brain but the functional connectivity within the brain dynamics. Here, we find lower dimensional states in ECoG data, looking for long-time scale patterns.

One of the biggest challenges we face is the complexity of the data; the brain data is high dimensional, and it contains structures across different length scales and different time scales. To analyze the data, we performed different dimensionality reduction methods to visualize brain dynamics in a lower dimensional state. Then, by looking at the embedding space from t-SNE, we found clustering features of embedding space are contiguous and also discrete. By looking at the amplitudes of all neural channels, we could group clusters on the embedding space together into different brain regions, with right hippocampus dentate gyrus (DG) being most dominant across the map. Then, we compare the transitions of the dataset to that of Markovian model generated from the data; we observe the dataset contains a much longer time scale, far beyond what Markov model can predict, indicating the presence of a complex dynamical structure that bridges scales.

Table of Contents

Introduction 1

Materials and Methods 4

    Data collection and pre-processing 4

    Wavelet transforms 4

    Spatial Embedding 6

    Training Set Generation and Re-embedding 7

    Markovian Models Generation 9

    Transition Matrices and Non-Markovian Time Scale 9

Results 9

    Embedding space dynamics 9

    Transition matrices and non-Markovian time scales 14

    Comparison with the Markovian model 16

Discussion 20

Table of Figures 21

References 22

About this Honors Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research field
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files