CCN meeting | David Alexander (KULeuven, Belgium)

17-01-2019 from 15:00 to 16:00
Henri Dunantlaan 2, room 4.4
Add to my calendar

Large-scale cortical travelling waves predict future localized cortical signals

Predicting future brain signal is highly sought-after, yet difficult to achieve. To predict future phase of neuronal population activity at localized recording sites, we exploit a dominant feature of cortex: its large-scale dynamics. Fourier analysis and principal components analysis (PCA) were used to construct from raw brain signal a model that, for each subject at each frequency of interest, predicts a non-overlapping portion of future signal. We analyzed ECoG data from 3 subjects and MEG data from 20 subjects, collected during a self-initiated motor task. The dominant eigenvectors of the PCA mapped large-scale patterns of past cortical phase to future ones, as smoothly changing waves over the whole measurement array. Test data yielded mean phase prediction errors as low as 0.5 radians at local sites, surpassing state-of-the-art methods of within-time-series methods or event-related models. Prediction accuracy was highest in delta to beta bands, depending on the subject, more accurate during high global power, but not strongly dependent on the time-course of the task. Prediction results did not require data from the to-be-predicted site. Rather, best accuracy depended on inclusion of very long wavelengths in the model. Large-scale, low spatial frequency traveling waves thus predict future phase activity at local sites.