Research Information System
NEURAL COMPUTING & APPLICATIONS, vol.30, pp.1341-1351, 2018 (SCI-Expanded)
https://link.springer.com/article/10.1007%2Fs00521-017-3006-8
ABSTRACT
Four asymmetry measurements (conventional coherence
function (CCF), cross wavelet correlation (CWC), phase lag index (PLI), and
mean phase coherence (MPC)) have been compared to each other for the first time
in order to recognize emotional states (pleasant (P), neutral (N), unpleasant
(UP)) from controls in EEG sub-bands (delta (0-4 Hz), theta (4-8 Hz), alpha
(8-16 Hz), beta (16-32 Hz), gamma (32-64 Hz)) mediated by affective pictures
from the International Affective Picture Archiving System (IAPS). Eight emotional
features, computed as hemispheric asymmetry between eight electrode pairs (Fp1
- Fp2, F7 - F8, F3 - F4, C3 - C4, T7 - T8, P7 - P8, P3 - P4, and O1 - O2), have
been classified by using data mining methods. Results show that
inter-hemispheric emotional functions are mostly mediated by gamma. The best
classification is provided by a neural network classifier, while the best
features are provided by CWC in time-scale domain due to non-stationary
nature of electroencephalographic (EEG) series. The highest asymmetry levels
are provided by pleasant pictures at mostly anterio-frontal (F3 - F4) and
central (C3 - C4) electrode pairs in gamma. Inter-hemispheric asymmetry levels
are changed by each emotional state at all lobes. In conclusion, we can state
the followings: (1) Nonlinear and wavelet transform-based methods are more
suitable for characterization of EEG; (2) The highest difference in hemispheric
asymmetry was observed among emotional states in gamma; (3) Cortical emotional
functions are not region-specific, since all lobes are effected by emotional
stimuli at different levels; and (4) Pleasant stimuli can strongly mediate the
brain in comparison to unpleasant and neutral stimuli.
DISCUSSION AND CONCLUSION
The results showed that the highest
performance was obtained for eight electrode pairs, which were placed on two
hemispheres symmetrically (Fp2 − Fp1, F3 − F4, C3
− C4, P3 − P4, O1 − O2, F7
− F8, T3 − T4, T5 − T6),
when CWC was applied to gamma of non-averaged emotional EEG oscillations of 6
s. Our findings were listed as follows: (1) The best phase synchronization
measurement was CWC; (2) Gamma was found to be highly sensitive to emotional
states (P, N, and UP); (3) The highest brain activation was induced by pleasant
pictures rather than other pictures in either N or UP; (4) The highest phase
coherence values produced by gamma were observed at anterio-frontal (F3
− F4) and central (C3 − C4) regions; (5) Two
contrasting emotional states (P and UP) were distinguished from each other
clearly, whereas the emotional state of UP was not clearly distinguished from
the state of N based on both phase synchronization level and inter-trial
variability of the phase difference in gamma; (6) The level of hemispheric
phase synchronization was changed by the state of emotion over the whole scalp;
and (7) The most useful data mining method was MultilayerPerceptron which is a
NN classifier.
In conclusion, we can
propose the use of CWC to extract emotional features from high-frequency
sub-band of non-averaged EEG oscillations. Considering two non-averaged EEG
oscillations, mediated by affective pictures and recorded from right and left
hemispheric regions, MPC and PLI can measure inter-trial variability of the
phase difference, while CWC can measure phase synchronization level. Results
show that high performance for classification of emotional states can be
obtained when the level of phase synchronization is used to estimate strong features
instead of the possible inter-trial variability of phase difference. Comparing
CWC to CCF, we should consider the basic difference between FT and WT: It is
known that high resolution can not be obtained by using FT in both time and
frequency, whereas we can perfectly determine the useful resolution by using WT
in time-frequency space. Therefore, FT-based CCF can not quantify the level of
phase synchronization in comparison to WT-based CWC to investigate functional
brain connectivity induced by static and visual emotional stimuli of 6 s. EEG
oscillations can be considered as summation of inhibitory and excitatory
postsynaptic nerve action potentials, because intrinsic cell currents are
mostly activated by ionic membrane channels as a result of ongoing interactions
between the thalamus and cortex. Both firing rate of individual neurons and
dynamic interplay among neurons within the local neuronal assembly have the
crucial role in processing neural information [54,55,56]. Considering right and
left hemispheres, neural populations can oscillate simultaneously depending on
depolarization phase initiated by nerve cell [57]. The brain is capable
of selectively integrate several neural populations during cognitive tasks in
gamma [58, 62]. Gamma is also found to
be coherent with spatiotemporal magnetic field [59, 63, 64]. In recent years, gamma
has been frequently investigated to understand the mechanism of highly ordered
cognitive functions such as emotional processes due to its high sensitivity to
emotional stimuli by means of coherence [34, 60, 61, 66]. Our new results are
compatible with these studies. Gamma has also been found to be sensitive to
phase synchronization in detecting emotional responses of patients with
alexithymia [65]. In future work, we are
going to compare controls to patients with schizophrenia with respect to the
same experimental paradigm for emotion recognition based on both single-channel
EEG complexity estimations and two-channel inter-hemispheric EEG synchronization
estimations in phase.
In
conclusion, inter-hemispheric neuronal functions, generated by long-duration
affective stimuli, can be analyzed by examining CWC in gamma. The results
support that there is a strong relationship between gamma-specific
inter-neuronal functions and emotional states at almost each brain region,
while anterio-frontal and central regions show relatively higher hemispheric
asymmetry levels in gamma.