?
Разделение пространственных и электрических компонентов ритмической кортикальной активности: подход на основе динамического моделирования.
Traditional studies on cortical activity relied on the assumption of space-time separability and neural data representation as across-trial averages. However, both invasive and noninvasive recording techniques detected traveling waves of neural activity in different brain areas in various contexts, invalidating the space-time separability assumption and making the static sources phenomenon inapplicable. The classical approach to analyzing non-invasive EEG/MEG data does not account for local movements of the source and assumes that the changes in the sensor signals exclusively reflect the fluctuation of electrical activity. Meanwhile, a significant amount of EEG and MEG signal variance may come from the spatial alterations of geometric properties of active neuronal populations. New methods are needed that consider the dynamic nature of the source and separate the spatial and electrical components of its activity. We propose an approach that models the spatial component using a local forward model from multichannel MEG data and the rhythmic electrical component as a frequency-modulated process. We employ the Unscented Kalman Filter to solve the resulting non-linear estimation problem aimed at recovering the electrical and spatial components of the neuronal dynamics underlying the measured MEG data and compare our technique with classical methods. The proposed methodology can be used for more reliable detection of rhythm desynchronization, which may become useful in brain-computer interfaces and in neurocognitive experiments for detecting brain states and facilitating context-dependent interactions with the brain.
In book
