Loading...

Brain Connectivity Based on the MVAR Model and their Relationship to each other

Abbaskhah, Ahmad | 2022

227 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55399 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Shamsollahi, Mohammad Bagher
  7. Abstract:
  8. During the time that the simplest action (rest) of the human brain is active, and for integration and coordination of the brain, different parts of it are in connection with each other. This connection can be directional and directionless, which are called effective connectivity and functional connectivity, respectively. It is clear that effective connectivity shows brain function better than other connectivity due to its directionality.One of the most common ways to define effective connectivity is the use of the Multivariate Autoregressive (MVAR). The MVAR model provides the time Cause of different signals on each other, meaning that the influence of the past of a variable on other variables is expressed as a linear relationship. Also, MVAR model is created according to the coefficients and variance of the noise estimated by the Electroencephalogram (EEG). Using the MVAR model coefficients, effective connectivity is defined in the domain of time and frequency, as well as the use of variances of noise in single and two-variable states, effective connectivity in the domain of granger.The aim of this research is to find the relationship between effective brain connectivity based on the MVAR model. At first, we tried to propose an approximate relationship between some brain connectivity in the frequency domain, but it was not very accurate (less than 0.08 on average for each pixel). In the next step, we tried to estimate one category of connectivity from another category, this problem is practically a regression problem, and we used neural networks, especially deep neural networks, by using these networks, we were able to estimate the time domain, Granger and frequency domain from each other with appropriate accuracy (the error is of the order of 5e-4 for three-channel data and 2e-3 for seven-channel)
  9. Keywords:
  10. Brain Connectivity ; Deep Neural Networks ; Effective Connectivity ; Electroencphalogram Signal ; Functional Connectivity ; Frequency Domain ; Time-frequency Domain ; Multivariate Autoregressive Model (MVAR)

 Digital Object List

 Bookmark

No TOC