Application to biomedical signals - Practical Biomedical Signal Analysis Using MATLAB - page 120

The progress in estimation of directionality was connected with introduction of the
measures based on the Granger causality principle: DTF and PDC.
DTF was successfully used, e.g., for estimation of EEG transmission during
overnight sleep [Kaminski et al., 1997], for finding patterns of LFP propagation dur-
ing locomotion in behaving animals [Korzeniewska et al., 1997], in epiloptegene-
sis [Medvedev and Willoughby, 1999], for finding interhemispheric dependencies in
presleep/wake states [Bertini et al., 2007], for determination of directionality of EEG
synchronization in Alzheimer disease [Babiloni et al., 2009], assessing the functional
connectivity [Astolfi et al., 2005].
Comparison of the different methods of directionality estimation for the same ex-
perimental data, namely EEG recorded in the awake state eyes closed, is illustrated
in Figure 4.9 . It is known that the main sources of activity in such a case are located
in the posterior parts of cortex and some weaker sources occur more frontally.
One can see that the clearest picture indicating the localizations of the sources
of activity (back of the head and some frontal localizations) was revealed by DTF.
dDTF showed transmissions determined by the anatomical tracts. In the bivariate
case the picture was quite confusing; namely one can even observe the reversal of
propagation, namely from C3 to P3 instead of P3 to C3. The propagation structure
obtained by means of PDC shows rather the sinks not the sources, so the results are
more difficult to interpret, which is connected with the normalization of the measure
(Sect. 3.3.2.4).
Inspecting Figure 4.9 it is easy to see that in the case of scalp EEG it is not impor-
tant to use an estimator showing only direct (not cascade) flows. It is more important
which parts of the cortex are communicating than what is the exact “wiring” scheme.
Therefore for estimation of transmissions of EEG registered by scalp electrodes DTF
is recommended. dDTF should be used for the intracranial electrodes where the di-
rect connections between brain structures are important [Korzeniewska et al., 2003].
In Figure 4.10 the results of the study of activity propagation during overnight
sleep are shown. The obtained patterns of EEG transmissions averaged over nine
subjects show a high degree of consistency. The most complicated pattern is visi-
ble for wakefulness since subjects differ in exact locations of sources. This pattern
becomes more organized during sleep. The results have a plausible physiological
explanation [Kaminski et al., 1997].
DTF is based on the phase differences between channels. It has non-zero value
only when there is a phase difference between signals from different derivations. This
fact was confirmed in the simulation studies and by the observation that even strong
artifacts, but with no phase difference between channels, did not influence results
at all. Volume conduction is a zero phase propagation; therefore no phase difference
between channels is generated. Thus in theory volume conduction shouldn't have any
influence on DTF results. In practice, it may have some influence, e.g., increasing the
noise level. However, this influence is not critical; it is much less important than in
case of other methods.
In some publications before the application of DTF the transformation to the
source space was performed, e.g., [Astolfi et al., 2005]. We find this procedure un-
necessary, since the influence of volume conduction is not significant in the case