Results and discussion of mental fatigue and college athletes’ response monitoring
Descriptive statistics for correct rate and correct response time (Unit: ms) were performed for all time periods, and the means and standard deviations were shown in Table 1. Some studies using subjective perception, behavioral data, and HRV frequency domain data have confirmed that 0–15 min can be used as a fatigue-free time period, 46–60 min as a fatigue time period, and the last 15 min as a fatigue recovery time period . The focus here was on whether there were significant differences between these three time periods, and the data were statistically processed using repeated measures ANOVA.
The results showed a significant difference in correctness across all time periods, F(4, 56) = 2.592, p = 0.046, η2 = 0.156. Post hoc tests revealed that the fourth time period was significantly lower than the first time period, and the first time period was slightly higher than the last 15 min (i.e., the music adjustment time period). There was a significant difference in correct response time for all time periods, F(4, 56) = 2.931, p = 0.029, η2 = 0.173. Post hoc tests revealed that the fourth time period was significantly higher than the first time period and the first time period was slightly lower than the last 15 min. The behavioral data suggest that performance was better in the no mental fatigue time period than in the music adjustment time period, and that the music adjustment time period was better than the mental fatigue time period.
The ERN (in the case of Fz, indicated by the arrow above the horizontal axis) and early Pe (indicated by the arrow below the horizontal axis) waveforms of the three electrodes Fz, FCz, and Cz during the fatigue-free, fatigue, and adjustment period were shown in Fig. 2.
The magnitude of Fz, FCz and Cz electrodes in the three periods ERN were in order of the fatigue-free period, the adjustment period and the fatigue period, and the magnitude of FCz was the largest. The amplitude of early Pe was relatively not obvious in the three periods, and the fatigue-free period was basically comparable to the adjustment period and larger than the fatigue period. The maximum position was also not obvious, and the amplitude of FCz and Cz were comparable.
The maximum magnitude values (Unit: μV) of 0-100 ms for each subject were extracted, and the mean and standard deviation of ERN were shown in the table below. The maximum magnitude values of 200-300 ms for each subject were extracted, and the mean and standard deviation of early Pe were shown in Table 2.
A 3(time period)*3(electrode) two-factor repeated measures ANOVA on the mean amplitude of 0–100 ms (ERN) showed a significant main effect of time period, F(2, 28) = 16.062, p = 0.000, η2 = 0.534. The 45–60 min amplitude (− 4.002) was significantly smaller than the last 15 min (− 6.261), and the last 15 min was significantly smaller than 0–15 min (− 8.328). The main effect of electrodes was significant, F(2, 28) = 4.445, p = 0.046, η2 = 0.241. Fz (− 5.619) and Cz (− 6.133) were significantly smaller than FCz (− 6.839). The interaction of time period* electrode was significant, F(4, 56) = 4.461, p = 0.010, η2 = 0.242. Simple effects were found for electrodes at 0–15 min (F(2, 28) = 7.240, p = 0.003, η2 = 0.341) and the last 15 min (F(2, 28) = 4.080, p = 0.028, η2 = 0.226), and the time period effects were significant at Fz (F(2, 28) = 13.400, p = 0.000, η2 = 0.489), FCz (F(2, 28) = 17.080, p = 0.000, η2 = 0.550) and Cz (F(2, 28) = 16.780, p = 0.000, η2 = 0.545).
A two-factor repeated measures ANOVA with 3 (time period) * 3 (electrode) for the mean amplitude of 200–300 ms (early Pe) showed that the main effect of time period was not significant, F(2, 28) = 0.398, p = 0.675, η2 = 0.028. The main effect of electrode was significant, F(2, 28) = 7.447, p = 0.003, η2 = 0.347. Fz (6.231) was significantly smaller than FCz (7.119) and Cz (7.128). Time period* electrode interaction was not significant, F(4, 56) = 1.094, p = 0.368, η2 = 0.072.
Figure 3 shows the 2D topography of ERN and early Pe after group averaged, including the brain topographic distribution during 0–50 ms and 200–250 ms for three time periods, which can visually describe the brain discharge. Referring to some previous studies, these two analysis periods were chosen mainly to consider the timing of ERN and early Pe wave amplitude appearance. As shown in Fig. 2, the ERN appeared before 50 ms and the early Pe appeared before 250 ms. The 0–50 ms showed the ERN amplitude variation, and the shades of blue were used to indicate the amplitude magnitude. The 200–250 ms showed the early Pe amplitude variation, and the shades of red were used to indicate the amplitude magnitude. The following was the same.
Between 0 and 50 ms after the error, there was a significant negative shift in the prefrontal lobe, with the maximum value occurring in FCz, and the magnitude was in the order of fatigue-free period, adjustment period and fatigue period. Between 200–250 ms after the error, there was a significant positive shift in the prefrontal lobe, with FCz and Cz more obvious, and the magnitude was in the order of adjustment period, fatigue-free period and fatigue period.
From the behavioral data, there was a directional change, with the fatigue-free period slightly better than the (high correct rate and fast response time) music adjustment period, and the fatigue-free period significantly better than the fatigue period, indicating a decrease in cognitive control during the fatigue period. The music adjustment period was better than the fatigue period, but no significant difference. From the ERN data, fatigue-free period was significantly greater than fatigue period, which is consistent with the studies of Boksem et al., Kato et al. and Lorist et al. [11,12,13, 36], indicating that error monitoring is impaired under mental fatigue. Also, the fatigue-free period was significantly greater than the music adjustment period, and the music adjustment period was significantly greater than the fatigue period, suggesting that ERN is a more sensitive indicator in evaluating the effect of music relaxation. From the early Pe data, the main effect of time period was not significant (suggesting that the effect of fatigue is not significant), the main effect of electrode location was significant (more pronounced with FCz and Cz), and no directional changes similar to ERN were found. The reason for this can be explained using the following explanation. The study  had shown that Pe is a member of the P300 family and is associated with subjects’ perception of errors and with the updating of the background in the brain regarding errors. It had also been shown  that a certain amount of ERN is produced whether the error is perceived or not, that Pe is evident only when the subject perceives the error. ERN and Pe reflect different error monitoring processing activities, ERN emphasizes error detection, Pe emphasizes awareness of error. The results of the behavioral and ERN data verified that the ability to monitor response effects decreased under mental fatigue. Mental fatigue can manifest itself in physical, psychological and behavioral ways. Consistency in trends and levels of typical indicators in these three areas can demonstrate the success of the manipulation of mental fatigue. Although the ERN and Pe data explained well the effect of mental fatigue on monitoring, in order to move ERPs toward standardization and norms, the meaning of differences in ERN and Pe scores should be further researched . It rebounded during music adjustment. The reason is that music has a direct or indirect effect on the limbic system of the brain and the reticular formation of the brainstem, which regulate the internal organs and somatic functions of the body. At the same time, these results confirmed the reliability, validity and security of the data based on AI algorithms.
There are some issues to be noted in the experiment. Firstly, because this study involved both mental fatigue and ERN elicitation, experiments with twice the number of presented subjects were conducted to obtain a better waveform, which was a great waste of subject data, and therefore should be strengthened in both follow-up studies and data processing. Secondly, this study focused on two variables, psychological fatigue and response monitoring. Based on the mechanism of psychological fatigue and the factors influencing response monitoring, third variables such as attention, arousal and stress should be included for mediating or moderating effects. Thirdly, in this study, based on the generalization of previous research results [3, 12], three electrodes, FCz, Cz and Fz, were chosen. The brain wave situation at the electrode positions of CPz and Pz could be continued to be explored if possible. It can enrich the statistics of the study. Finally, regarding the search for other means of mental fatigue prevention and recovery. One of the subjects who did not successfully induce mental fatigue said that the first few minutes of the formal experiment, more mistakes were made, thinking that they could not be changed anyway. So he began to look for certain ways to cope with the situation, mainly with two elements. When performing the operation, according to the order of picture presentation, mentally meditating on 1, 2 and 3, which correspond to the gaze point, response stimulus and button respectively. This created the rhythm. He meditated on the number of consecutive correct, to see if there is a boost. There were more than 10, more than 20, the most time is 96. In this way, the more you do it, the easier it becomes. He did not experience mental fatigue. This case speculates that mental fatigue can be prevented and alleviated by developing a sense of rhythm and setting goals, which also requires algorithms to verify based on the idea of cross-fertilization of disciplines. In addition, recent researches have demonstrated that mindfulness , nature , and other modalities can also help in the recovery of response monitoring in such cases.
Results and discussion of burnout and college athletes’ response monitoring
To verify that the 15-min task did not significantly induce mental fatigue in the burnout group, subjective perception tests were conducted on 12 subjects. Repeated measures ANOVAs were conducted for difficulty, effort, and fatigue before and after the 15-min cognitive task. Their means and standard deviations were shown in Table 3.
In terms of direction, the burnout group was basically on the rise. There was no significant difference in difficulty scores before and after the cognitive task, F(1, 11) = 0.805, p = 0.389, η2 = 0.068. There was no significant difference in effort scores before and after the cognitive task, F(1, 11) = 1.941, p = 0.191, η2 = 0.150. Fatigue before and after the cognitive task scores were not significantly different, F(1, 11) = 3.143, p = 0.104, η2 = 0.222. The above results indicated that no mental fatigue occurred with the 15-min task.
A multivariate ANOVA was performed on the correct rate and correct response time (Unit: ms) of the two groups of subjects with and without burnout in college athletes, and the means and standard deviations of the two groups were shown in Table 4.
The results showed that the no burnout group had a higher correct rate than the burnout group, with no significant difference, F(1, 22) = 0.062, p = 0.806, η2 = 0.003. The no burnout group had a lower reaction time than the burnout group, with also no significant difference, F(1, 22) = 0.667, p = 0.423, η2 = 0.029. These indicated that the no burnout group’s behavioral performance was better than that of the burnout group.
The waveforms of the three electrodes Fz, FCz, and Cz in college athletes with and without burnout were shown in Fig. 4.
The maximum magnitude (Unit: µV) of 0–100 ms for each subject was extracted, and the mean and standard deviation of ERN for 12 subjects were shown in the following table. The maximum magnitude of 200–300 ms for each subject was extracted, and the mean and standard deviation of early Pe were shown in Table 5.
A 2(group)*3(electrode) two-factor repeated measures ANOVA was performed on the mean amplitude of 0-100 ms. The results showed a margin significant main effect for group, F(1, 11) = 4.748, p = 0.052, η2 = 0.301. The amplitude of the burnout group (− 6.547) was smaller than that of the no burnout (− 9.297). The main effect of electrodes was significant, F(2, 22) = 17.552, p = 0.001, η2 = 0.615. Fz (− 7.081) was significantly smaller than Cz (− 7.968) and Cz was significantly smaller than FCz (− 8.717). Group* electrode interaction was not significant, F(2, 22) = 0.845, p = 0.443, η2 = 0.071.
A 2(group)*3(electrode) two-factor repeated measures ANOVA was also performed on the mean amplitude of 200-300 ms. The results showed that the main effect of group was not significant, F(1, 11) = 0.025, p = 0.878, η2 = 0.002. The amplitude of the burnout group (8.629) was slightly smaller than that of the no burnout (8.878). The main effect of electrodes was significant, F(2, 22) = 16.573, p = 0.000, η2 = 0.601. Fz (7.689) was significantly smaller than Cz (9.239) and FCz (9.332). Group* electrode interaction was not significant, F(2, 22) = 0.123, p = 0.751, η2 = 0.011.
Figure 5 showed the 2D topography of ERN and early Pe after group averaging, including the distribution of brain topography at 0–50 ms and 200–250 ms with and without burnout. Between 0 and 50 ms after the error, there was a significant negative shift in the prefrontal lobe (FCz was most pronounced), and the degree was greater in the group without burnout than in the group with burnout. Between 200 and 250 ms after the error, there was a significant positive shift in the prefrontal lobe (FCz and Cz were more pronounced), and the degree was slightly greater in the no burnout group than in the burnout group.
From the behavioral data, the group without burnout was better than the group with burnout, but the difference was not significant, consistent with the previous study . From the ERP data, the ERN index in the no burnout group was greater than that in the burnout group, and the difference was borderline significant, indicating that the ability to monitor response effects is impaired under burnout and that ERN has some sensitivity in evaluating burnout in college athletes. The maximum electrode position is at FCz, which is consistent with ERN under mental fatigue in the first experiment. Regarding the early Pe index, there was no significant difference between the two groups with and without burnout. The maximum electrode location was at FCz and Cz, which is consistent with the case of early Pe under mental fatigue. When doing the 2D Cartoon chart, the time period used was 200–250 ms. In fact, most of the early Pe in this experiment appeared between 250 and 300 ms. But for two reasons the choice was still chosen to be 200–250 ms, one was to correspond to the preceding and following text, and the other was that the waveforms are basically consistent with the graphical pattern of 250–300 ms. As a result, the ability to monitor response effects decreases under burnout. The reasons for the absence of significant differences in results may be related to the mechanisms by which burnout occurs. Burnout is related to factors such as personality, sense of self-control, motivation and social support. Burnout is also more a symptom of mood changes than of changes in cognitive functioning. As a result, individuals do not show reduced levels of behavior and cognition, but are simply less willing to complete the task in question.
In addition, the collection and analysis of data need to be strengthened. The experimental paradigm should be improved, and the experimental data should be further processed using other AI algorithms. Firstly, the creation of computer data simulator. A study had been conducted to build a simulator of visual N2/N2pc event-related potential components in order to assess the accuracy of estimates . Secondly, the design of brain-computer interfaces. ERN and Pe can be widely used for neurorehabilitation of different populations [44, 45]. Error-related potential-based brain-computer interfaces have become a hot topic of research in this area. Generally, ERN and Pe are obtained by stacking several times. The single detection of ERN and Pe is a technical bottleneck in their application to brain-computer interface systems. The advantages of the wavelet transform algorithm will continue to be exploited. Therefore, researches in brain science have provided new inspiration for the development of devices and the evolution of artificial intelligence algorithms.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.