Discussion

In this study, we recorded and analyzed un-averaged single-trial EEG data in 31 driving experiments from 10 volunteer drivers under two different driving conditions -- motion and motionless. The hexapod motion platform that simulated driving events allowed us to study neural correlates of kinesthetic stimuli, which is difficult, if possible, to study in regular EEG laboratories. We performed ICA to separate the EEG contributions of distinct brain processes to explore their individual and joint event-related dynamics following Stop-Go and deviation events through ERP differences and time-frequency analysis (ERSP). The 9 independent component clusters here identified by their similar scalp projections and activity spectra resemble classes of EEG phenomena long described by neurologists from observations of paper data displays such as central and lateral alpha, left and right mu, and frontal-midline theta rhythms. Alpha power of the mu component cluster was strongly blocked (~-5dB) around the peak of platform movement in Motion-Stop and Motion-Go events. A sharp negative was found in the central midline component cluster only in

Motion-Deviation events. We believe that these two features were induced by kinesthetic stimuli.

3.4.1 Phenomenon in Mu Component

Mu rhythm (μ rhythm) is an EEG rhythm recorded usually from the motor cortex of the dominant hemisphere. It is also called arciform rhythm given the shape of the waveforms. It is a variant of normality, and it can be suppressed by a simple motor activity such as clenching the fist of the contra lateral side, or passively moved (Thilo et al., 2003; Loose et al., 2002; Parker et al., 2001). Mu is believed to be the electrical output of the synchronization of large portions of pyramidal neurons of the motor cortex which control the hand and arm movement when it is inactive.

Deviation events involved subject responses to steer the vehicle back to the cruising position, thus it is expected that mu power would be blocked following deviation events. Our results also showed unexpected strong mu blocking in response to Motion-Stop and Motion-Go events in which no action was involved, suggesting kinesthetic stimuli could also induce mu blocking. Following deviation events, mu power was strongly blocked in both motion and motionless conditions (cf. Figures 3-9 and 3-10). Mean subject RT indexed by the first steering action in response to Motion-Deviation events leads that in response to Motionless-Deviation events by about 50 ms. Thus, we expect that the latency of mu blocking in Motion-Deviation events would lead that in Motionless-Deviation events by a comparable length.

However, Figure 3-10 reveals that the mu-blocking latency discrepancy between the two conditions is about 250~300 ms, which could not be attributed entirely to the subject RT latency difference. Mu blocking thus appears associated with kinesthetic stimuli delivered to the drivers. In short, long-lasting mu blocking following deviation events began with the EEG brain dynamics induced by kinesthetic stimuli, followed

by marked mu power decrease associated with subject motor actions. Table 3-1 gives us the information that response time in Motion-Deviation events was only 50 ms faster than in Motionless-Deviation events. By these two results we discovered 200~250 ms duration which was not related to steering action.

3.4.2 Phenomenon in Central Midline Component

The central midline component cluster exhibits a sharp negativity in averaged ERP following Motion-Deviations, but the negativity is missing from the ERP following Motionless-Deviations. The mean ERP in deviate-to-right and deviate-to-left conditions was almost identical. ERP images also show a weaker negative ERP time-locked to subjects’ reactions (the black line in the ERP image), which again is comparable following Motion-Deviation and Motionless-Deviation events. Response time in Motion-Deviation events was approximately 50 ms shorter than that in Motionless-Deviation events (as shown in Table 3-1), consistent with a previous report (Wexler et al., 2001) which showed that the absence of motion information increased response times to external movement perturbations.

The sharp negativity in the ERP of the central midline component cluster is also consistent with previous VESTEP studies of Elidan et al. (1982, 1984, & 1987). They showed a negative potential near Cz or forehead, induced by external kinesthetic stimulus. They did not, however, report any mu blocking in response to the kinesthetic stimuli, which to the best of our knowledge has never been reported in the past. The reason is due at least in part to the fact that our experimental environment, which combined visual and vestibular interaction and driver response, was more complicated and realistic than the experimental setups used in previous studies.

3.4.3 EEG Alpha Activity related to Drowsiness

Traditionally, EEG alpha band was used as an indicator of drowsiness estimation during driving (Lin et al., 2005; Eoh et al., 2005). Alpha power has been reported to index the level of drowsiness in attention-sustained experiments in a laboratory setting. In this study, our results showed that alpha-band activity varies during driving, especially when the vehicle is moving and delivers kinesthetic stimuli to the drivers and passengers, which might confound the fatigue-related alpha power changes in driving. Thus, more care must be taken to examine the validity of using alpha power to index drowsiness level in real driving.

Our experiment results show that kinesthetic stimulus during driving induces (1) Mu blocking in the somatomotor components, and (2) Sharp negative ERP in central midline components. The mu blocking appeared to be induced by two types of stimuli successively. When the subjects received kinesthetic inputs, their alpha activities in the left and right mu components were blocked. After a short period, the subjects made an adjustment to balance them, inducing a secondary mu blocking. The alpha power variation induced by the motion of the vehicle might interfere with the estimation of the driving cognitive state based on the fluctuations in the alpha power spectra.

Furthermore, negative ERP was found in central midline components following kinesthetic stimulus onsets. These results demonstrate that multiple cortical EEG sources respond to driving events distinctively in dynamic and static/laboratory environments. A static driving simulator could not induce some cognitive responses that are actively involved in real driving. We also reported that the absence of driving motion will increase the reaction time to external perturbations by studying the response time in deviation events. Thus a driving simulator with a motion platform is crucial to studying event-related brain activities involved in real driving.