Oscillatory brain rhythms are considered to originate from synchronous synaptic activities of a large number of neurons . Synchronization of neural networks may reflect integration of information processing, and such synchronization processes can be evaluated using MEG time-frequency analyses . Aberrant resting-state oscillatory activities have been observed in various pathological conditions [15, 24]. We found that, after the 0-back test, decreased alpha power was shown in the angular gyrus and increased levels were shown in the temporal, postcentral, and frontal gyrus. In contrast, after the 2-back test, decreased alpha power was shown in the frontal gyrus and increased levels were shown in the parietal lobule, parahippocampal gyrus, uncus, postcentral gyrus, and frontal gyrus. For beta power, after the 0-back test, increased power was shown in the temporal, frontal, cingulate, and precentral gyrus. After the 2-back test, decreased power was shown in the frontal gyrus and increased power was shown in the temporal gyrus and parietal lobule. In addition, the beta power change in the left precentral gyrus was negatively associated with task performance of the 0-back test session; and the alpha power changes in the right middle and superior frontal gyrus were negatively associated with the task performance of 2-back test session.
In our previous EEG study, beta power densities on the P3, Pz, and O1 electrodes and alpha power densities on the P3, Pz, O1, and O2 electrodes were decreased after a fatigue-inducing 2-back test session, while alpha power densities on the P3 and Pz electrodes were decreased after a fatigue-inducing 0-back test session. Although there are some similarities between the results of the EEG and MEG studies, the differences are apparent. EEG measures electrical fields, which are based on the difference in potentials between an EEG electrode and a reference electrode. In addition, the EEG signal is influenced by the electrical conductivity of intervening tissues such as the skull. In contrast, MEG measures magnetic fields, which does not use a reference, and MEG provides superior spatial resolution and signal-to-noise ratio relative to EEG. Therefore, MEG could identify accurate changes of the mental fatigue-related oscillatory brain activities that were not evident in our previous EEG study.
Multiple, broadly distributed, and continuously interacting dynamic neural networks are achievable through the synchronization of oscillations at particular time-frequency bands. Alpha, one of the large-scale rhythmic oscillations in the brain, is generated in the process of interactions between thalamocortical neurons and GABAergic (γ-aminobutyric acid) cells in the thalamic reticular nucleus [25, 26]. This time-frequency band is related to complex cognitive processes such as attention, memory, and mental imagery [27–29]. Chaudhuri and Behan proposed that the fatigue is related to the activation of the thalamo-frontal feedback loops [5, 28, 30], and the overactivation of the thalamo-frontal feedback loops in order to compensate for the functional loss caused by fatigue has been reported in fatigue with some diseases such as multiple sclerosis [31–33] and chronic fatigue syndrome [34, 35]. Since mental fatigue induced by 2-back trials suppressed spontaneous alpha power, i.e., desynchronization due to intrinsic events, in the frontal gyrus, this desyncronization might have some relationships with the activation of the thalamo-frontal feedback loops.
The motor cortex exhibits resting-state synchronization of the oscillations at beta frequency band [36, 37]. It has been demonstrated that these oscillations appear to be under the direct control of GABAergic modulation [38, 39]. These oscillations are facilitated by increasing the inhibitory drive of GABAergic interneurons via GABA-A receptors, which lengthens the inhibitory post-synaptic potential decay time, thereby reducing the frequency of the locally oscillating neuronal network population. Consequently, this serves to facilitate the recruitment of principal cells to the oscillating population, giving rise to an increase in the amplitude of the oscillatory power, as the participating neuronal pool is increased . In contrast with the results of the 2-back test, the beta power in the left precentral gyrus was increased after 0-back test trials. Interestingly, the increased power in the left precentral gyrus was negatively associated with the percentage correct and positively associated with the reaction time of the 0-back test session. Inhibitory mechanisms apparently suppressed the beta power in the motor areas according to the level of impaired task performance, i.e., mental fatigue. Fatigue is a bio-alarm that senses risks, warns the organism, and orders rest. Therefore, some inhibitory mechanisms that order rest to in order to avoid overwork and even homeostatic catastrophe may exist, and we may recognize that the inhibition is caused by fatigue, resulting in a sensation of fatigue . The increase in inhibitory input to the motor cortex with physical fatigue has been suggested in behavioral  and MEG  studies; enhanced and persistent activation of the inhibitory system may be related to the pathophysiology of chronic fatigue [44, 45]. Since this inhibitory system alone causes impaired task performance or even cancellation of task trials, some compensatory mechanisms would be necessary to maintain the task performance. The 0-back test did not require a higher level of mental load; thus, compensatory mechanisms may not be necessary to maintain task performance during the task trials, overactivation of the inhibitory system seems to play a central role in the pathophysiology of mental fatigue caused by 0-back test trials, in contrast to the mental fatigue caused by 2-back test trials.
While the results of the present study are suggestive of the mechanisms of mental fatigue, only a limited number of participants and limited time-frequency bands were tested. To generalize our results, studies involving a larger number of participants and a variety of time-frequency bands will be needed. In addition, we did not measure the objective markers of mental fatigue in urine and serum  and did not investigate the relationship between the MEG data and these markers. Furthermore, although the participants did not complain of physical fatigue, we could not exclude the possibility that the MEG data was affected by physical fatigue. Finally, the use of the task with more burdens, such as 3-back or more, longer duration of tasks could effectively extract the fatigue as the difference of these two conditions. We used 30-min 2-back test as a fatigue-inducing task based on the results of our previous study : To evaluate mental fatigue, participants completed advanced trail-making test (ATMT ) for 30 min, and after the 2-back test session for 30 min, ATMT performance was impaired, and this session was shown to be mental fatigue-inducing tasks. In the ATMT, circles numbered from 1 to 25 were randomly located on the display of a personal computer, and the participants were required to use a computer mouse to click the center of the circles in sequence, starting with circle number 1. The task performance was evaluated by number of errors. In addition, by using the 0-back and 2-back tasks, we could extract the fatigue as the difference related to the MEG responses to visual stimuli of these two conditions [47, 48]. Therefore, we adopted 30-min 2-back test as a fatigue-inducing task.