Prípravný seminár s názvom Tensor decomposition analysis of EEG signals during BCI-assisted stroke rehabilitation

Pozývame vás na prípravný seminár MSc. Samana Seifpoura k obhajobe dizertačnej práce na tému:

 

Tensor decomposition analysis of EEG signals during BCI-assisted stroke rehabilitation

Dátum: Štvrtok, 09.06.2022 o 10:00
Miesto konania: Zasadacia miestnosť ÚM SAV, v. v. i.

Abstrakt:

Tensor decomposition analysis of EEG signals during BCI-assisted stroke rehabilitation

By: Saman Seifpour

Supervisor: Prof. Ing. Alexander Šatka, CSc.

Rehabilitative technologies based on brain-computer interface (BCI) show promising clinical results in the functional recovery of post-stroke patients. The high dimensionality of BCI-EEG data, especially in a longitudinal motor-training paradigm, is a real challenge for the analysis. Most signal processing frameworks investigating BCI training effects on neural modulations are univariate. While, due to neuroplasticity characteristic that enables reorganization of neural pathways in the brain, it can be hypothesized that specific brain networks can be activated after each MI BCI training session. These networks not only have a specific spatial distribution but also have specific spectral signatures that differ based on cortical locations, experimental conditions, recording, and tasks. In this dissertation thesis, to contribute to a deeper understanding of the neural mechanisms underlying motor-related behaviors, we employ a comprehensive analytical framework based on parallel factor analysis (PARAFAC). Using this framework, we identified a set of lateralized narrow-band subject-specific sensorimotor rhythms by projecting the dominant PARAFAC spatial and spectral weights to the oscillatory part of the EEG spectrum. In this presentation, we will show how the identified narrow-band oscillatory rhythms are contributed to sensorimotor repair mechanisms. We will qualitatively show observed interplays between rhythms, focusing on the mu-rhythm dynamical modulation and its relation to lower and higher beta sensorimotor rhythms. We will discuss functional distinctions between these rhythms’ modulations during different phases of MI depending on whether the eyes were open or closed. We will discuss the PARAFAC potential to precisely identify the main profiles of neural oscillations induced by MI and provide an informative representation of the high dimensional BCI-EEG data, specifically when recorded in a longitudinal design.