Content area
Objective. The assessment of mental fatigue (MF) and attention span in educational and healthcare settings frequently relies on subjective scales or methods such as induced-task interruption tools. However, these approaches are deficient in real-time evaluation and dynamic definitions. To address this gap, this paper proposes a Continuous Quantitative Scale (CQS) that allows for the natural and real-time measurement of MF based on group-synchronized electroencephalogram (EEG) data. Approach. In this study, computational psychophysiology was used to measure MF scores during a realistic class. Our methodology continuously monitored participants’ psychological states without interrupting their regular routines, providing an objective evaluation. By analyzing multi-subject brain-computer interface (mBCI) data with a collaborative computing approach, the group-synchronized data were obtained from 10 healthy participants to assess MF levels. Each participant wore an EEG headset for only 10 min of preparation before performing a sustained task for 80 min. Main results. Our findings indicate that a lecture duration of 18.9 min is most effective, while a duration of 43.1 min leads to heightened MF levels. By focusing on the group-level simultaneous data analysis, the effects of individual variability were mitigated and the efficiency of cognitive computing was improved. From the perspective of a neurocomputational measure, these results confirm previous research. Significance. The proposed CQS provides a reliable, objective, memory- and emotion-free approach to the assessment of MF and attention span. These findings have significant implications not only for education, but also for the study of group cognitive mechanisms and for improving the quality of mental healthcare.
Introduction
The “common knowledge” of a 15-min attention span in educational settings, while commonly referenced, is underpinned by a body of research (Schwartz et al. 2019; Ali et al. 2022; Bradbury 2016) on the dynamics of student attention and mental fatigue (MF). Collectively, these studies suggest that attention tends to initially increase during the first 15 min of a lecture, followed by a subsequent decline in attention and an increase in MF. The foundational work (Hartley and Davies 1978) solidified the 10–15 min benchmark for attention span, which has been supported and extended by subsequent researches (Maddox and Hoole 1975; Kleinsasser and McKeachie 1994; Wilson and Korn 2007; Szpunar et al. 2013; Vekaria and Peverly 2018; Argyriou et al. 2022) in relation to the declining attention span and note taking. Despite these findings, there is a recognized challenge in accurately measuring MF from population neuroscience, due to the subjective nature and the methodological limitations of the assessment tools available at the time.
MF not only has a broader impact on physical performance (Sun et al. 2021; Habay et al. 2021), but also affects cognitive functions such as learning-related attention and reaction time (Chen et al. 2023; Kikuchi et al. 2020; Sun et al. 2021)and has a negative impact on cognitive performance, such as the ability of drivers (Zhang et al. 2021) and pilots (Hamann and Carstengerdes 2023; Wingelaar-Jagt et al. 2021). In severe cases, MF can contribute to mental disorders such as depression and anxiety (Nagasaki et al. 2022). It is well known that there are no methods to directly quantify MF states (Eronen and Bringmann 2021). Furthermore, the recent phenomena of “zoom fatigue” (Knox et al. 2023; Nesher Shoshan and Wehrt 2021) and the educational disruptions caused by the COVID-19 pandemic (Atwa et al. 2022)underscore the urgent need for direct, reliable, real-time MF assessment scales. However, measuring MF is challenging, as it is influenced by many factors, such as task characteristics, individual differences, environmental conditions, and emotional states. Existing approaches to the assessment of MF can be divided into subjective, behavioral, and physiological methods. Subjective methods typically rely on self-report scales or questionnaires (Gopi and Madan 2023), such as the Multidimensional Fatigue Inventory-20 (MFI) (Smets et al. 1995) and the Fatigue Severity Scale (Lerdal 2021). These methods are easy to administer and widely used, but there are several limitations, such as recall bias, social desirability, lack of real-time feedback, and inability to capture dynamic changes in MF. Behavioral methods include performance indicators or cognitive tests (Zorowitz S, Niv Y 2023), such as the Psychomotor Vigilance Task (PVT). Physiological methods (Masri et al. 2023) rely on biological signals, biomarkers or multimodality integration, such as electrooculography (EOG) (Zhang et al. 2022), electrocardiogram (ECG) (Huang et al. 2018), electroencephalogram (EEG) (Hamann and Carstengerdes 2023; Tian et al. 2018). However, these objective measures also have drawbacks, such as task interference, learning effects, motivation effects, and lack of ecological validity (collection of single-subject data per trial or requiring long preparation time), always require participants to carry bulky equipment or to be monitored before and after the experiment in a non-natural state. In addition, these approaches cannot continuously assess cognitive states.
To overcome these limitations, a novel scale for MF measurement based on group-synchronized EEG data was proposed. Our method is based on the concept of computational psychophysiology (CPP), which is defined as the use of computational methods to analyze and interpret physiological signals for the quantitative understanding of human psychological states and processes (Hu et al. 2022).
In this study, CPP was used to objectively measure real-time cumulative MF during a sustained 90-min naturalistic class. Using data-driven collaborative computation and a synchronized system, the effective attention span (18.9 min) and the optimal class duration (43.1 min) were quantitatively proven. The main innovations of this study are the development of a Continuous Quantitative Scale (CQS), which allows for a natural and real-time measurement of MF based on group-synchronized EEG data, and 18.9 min effective attention span from a neurocomputational point of view, thus confirming previous studies. The CQS is derived from the analysis of the EEG power spectrum and the EEG coherence among the subjects. It can capture the dynamic changes in the MF levels during realistic and sustained tasks, such as attending a class. The CQS is independent of memory and emotion, which are often confounding factors in subjective and behavioral methods.
The rest of the paper is organized as follows. Sect. “Methods” is ordered to the key work on MF measurement methods. Sect. “Results and discussion” presents the experimental results, findings, discussions, comparison and future work. Sect. “Conclusion” concludes the paper.
Methods
Both paradigm design and data collection are crucial steps in implementing a psychophysiological model to solve the CPP inverse problem. This is because physiological signals are not specific to psychological variables, and inverse validation of the paradigm design requires appropriate experimental conditions and the selection of appropriate direct features.
The EEG is widely considered as the “gold standard” for MF detection (Borghini et al. 2014), as the signals effectively and directly monitor functional and physiological changes in brain activity. As noted in research (Zhao et al. 2012; Fairclough et al. 2005; Fairclough and Venables 2006), when a person experiences MF while awake, there is a significant increase in alpha and theta power, while beta power decreases significantly. Wei (Wei et al. 2018) successfully detected MF in drivers with over 80% accuracy using the energy of alpha, beta, and theta band energy as features. Therefore, in this study, both EEG signals and their associated rhythms were selected as direct features for MF detection.
Feature computation
A method derived from the classical inverse problem of mathematical physics was used to compute the features. This knowledge-based inverse problem consisted of two steps. In the first step, two well-known formulae were chosen as evaluation methods. In the second step, an experimental paradigm was designed and the group-synchronized data were used to perform an inverse attention span or MF computation. In the natural condition, subjects were not explicitly informed about the experimental procedure. They simply wore the EEG caps and went through the three periods. An inter-correlation analysis between the group cognition resulted from each section. The mutual paired t-tests were used to analyze the group data and to generate a fitted curve. This curve can potentially be used as an assessment criterion for attention span in future studies.
Previous research (Balandong et al. 2018) has shown that MF is associated with increased power in the theta () and parietal alpha () EEG rhythms. Sleepiness is often indicated by increased theta and alpha activity, accompanied by a decrease in the beta band (Ko et al. 2017). EEG power spectrum features are commonly used in cognitive science to assess MF or attention span (Eoh et al. 2005). Brain rhythms can generally be divided into five sub-bands: : 0.5–4 Hz; : 4–8 Hz; : 8–13 Hz; : 13–25 Hz and : 25–40 Hz. A classical formula is represented as follows:
1
The other classical formula is (da Silveira et al. 2016):
2
where the total band power is:3
where represents the result of the signal at a given frequency . F1 values are used as an indicator to assess MF levels, with higher F1 values indicating higher MF levels (Ko et al. 2017). Conversely, F2 values show a negative correlation with MF levels, with higher F2 values indicating lower MF levels (Eoh et al. 2005).Participants
The study involved 10 healthy subjects who were students and staff members of the university. All participants provided signed consent forms. The experimental protocols were approved by the Institution Review Board of Tsinghua University.
Group cognition experimental setting
To conduct the experiment in a natural way and to calculate the MF collaboratively, a questionnaire-free experiment was designed and conducted. For tasks involving more than 10 subjects, especially in real classroom settings, it is critical to use quick-wear and compact devices that provide reliable signals (Huang et al. 2023). Prior to the experiment, all participants wore EEG-caps for 10 min (refer to the Prepare as shown in Fig. 1A). The experiment was divided into three periods: Period 1—Guidance (2 min), Period 2—Online teaching (27 min), and Period 3—Offline teaching (49 min), as shown in Fig. 1A. The procedure lasted 80 min without any breaks, from 10:00 a.m. to 11:20 a.m. Before the teaching, the resting EEG data were measured during the 2-min Guidance session (1–0-Contrast-G) as the baseline. During Period 2, the experimental protocol included 2 min of eye closure (1–1-close), 2 min of eye opening (1–1-open), 10 min of watching the teaching video at normal speed (1–1-oT10), 5 min of watching the video at double speed (1–2-oT20), 6 min of watching the video at 1.5 times the speed (1–3-oT15), and finally 2 min of watching the video at normal speed (1–4-oT10). Period 3 consisted of 2 min of eye opening (2–0-open), 2 min of eye closing (2–0-close), and 45 min of offline teaching (2–1-1).
Fig. 1 [Images not available. See PDF.]
The process of cognitive computation. A The experimental setting, B Temporal-spatial diagram, C EEG preprocessing and feature computation
Multi-subject BCI recordings
Accurate continuous quantitative assessment of MF requires that the acquired multi-subject, multi-channel EEG signals represent the same psychological semantics. The approach used to acquire the data determines whether the psychological state can be effectively inferred from the mapping of physiological responses to psychological variables. A multi-subject brain-computer interface (mBCI) system was used to obtain group-synchronized EEG signals from the occipital region under noisy environmental conditions. Each subject’s EEG was recorded by an individual amplifier (BLueBCI Inc.) connected via dry electrodes. The ten amplifiers (A1-A10, which also represented to 10 subjects) were synchronized by the same trigger transmitter (Huang et al. 2023). A custom software recorder was used to record the data. Each EEG-cap was used with 10 dry Ag/AgCl electrodes placed according to the 10–20 system, with electrode placements including PO6, PO4, POz, PO3, PO5, O1, Oz, O2. A reference and a ground electrode were placed on the forehead. EEG data were recorded at a sampling rate of 1,000 Hz.
EEG preprocessing
Prior to the preprocessing, the reliability of the EEG was first assessed using the two criteria from the view of hardware characteristics and the recorded video, respectively. For the hardware characteristics, the common mode rejection ratio (CMRR) and the noise of the amplifiers can be identified. For the recorded video, the malfunctioning EEG caps and the abnormal behavior of the subjects can be captured. Therefore, the raw data of subjects A4 and A9 were excluded due to abnormal noise levels (more than 5 μVrms) or low CMRR (less than 100 dB). The following acquired data were preprocessed through several steps, including segmentation, data filtering, and feature calculation, as shown in Fig. 1C.
Firstly, the lecture segments of the procedure were further sliced with similar time intervals corresponding to different time periods. For example, the “1-1-oT10” segment was sliced into every 5-min sections (“1–1-oT10-1”, “1–1-oT10-2”) without overlap, and the first 20 min of the 45-min offline teaching segment was similarly sliced (“2–1-1–1”, “2–1-1–2 ”, “2–1-1-3”, “2–1-1–4”). For the long-term sustained class, the procedure was further segmented to produce a more balanced, aligned data for better analysis, visualization, and scaling.
Secondly, the collected data were aligned to the lecture procedure in time, including 14 segments, “1–0-Contrast-G” (2 min), “1–1-close” (2 min), “1–1-open” (2 min), “1–1-oT10-1” (5 min), “1–1-oT10-2” (5 min), “1–2-oT20” (5 min), “1–3-oT15” (6 min), “1–4-oT10” (2 min), “2–0-open” (2 min), “2–0-close” (2 min), “2–1-1–1” (5 min), “2–1-1–2” (5 min), “2–1-1–3” (5 min), “2–1-1–4” (5 min). Due to the poor data from A4 and A9, the 896 epochs were finally collected in this study. The 64 epochs (EEG data during the “1–0-Contrast-G” segment) were considered as the baseline. The EEG data during the “1–1-close”, “2–0-close” segments were used to check the validity of the data.
Thirdly, all segmented epochs were filtered using a bandpass filter with a frequency range of 0.5–80 Hz to remove the DC offset and the high frequency noise. Any channel data identified as being of poor quality (based on careful visual inspection of the power spectra and time domain) were excluded. Specifically, the epochs with (1) an amplitude greater than 100 μV, (2) too-rapid discontinuous changes, or (3) continuous the same for more than 50 points, were considered as bad data and removed from the following analyses. All epochs were then filtered with a notch filter at 50 Hz, and then down-sampled to 250 Hz.
Fourthly, to further reduce muscle artifacts and ambient noise a Chebyshev type I bandpass filter was used, with a passband of [2, 65] Hz and stopband edge frequencies of [1, 75] Hz, along with < 3 dB ripple in the passband and 40 dB attenuation in the stopbands. Independent Component Analysis (ICA) was used to detect other artifacts (e.g., EOG, ECG), and components predominantly containing these artifacts were removed. For the excluded epochs, interpolation was then performed on the invalid channels using EEGLAB. After that, the revised 896 epochs remained for further analysis.
Finally, the data were transformed into the frequency domain using the Fast Fourier Transform (FFT) algorithm. The EEG power spectrum and the EEG coherence were calculated for each epoch, channel, and frequency band. The frequency bands were delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–25 Hz), and gamma (25–40 Hz). The above preprocessing was applied to all epochs.
Mutual T-test matrix and comparison
To identify the boundary threshold of MF levels or attention span, regression models were calculated using synchronized data. The statistical explanation of the common knowledge “approximately 15 and 45 min” (Bradbury 2016) was performed using the mutual t-test matrix and formulae (see the Results and Discussion section).
Results and discussion
Experimental results
During the experiment, contrast data were collected from three sections: “1–0-Contrast-G”, “1–1-close”, and “2–0-close”. Particularly, Fig. 1B highlights that the FFT and the temporal domain were prominent during the “1–1-close” and “2–0-close” periods. The MF level at the beginning of the experiment during the “1–0-Contrast-G” period was used as the baseline for subsequent teaching sessions. These results validated the use of raw data obtained from the quick wearable mBCI system.
Data from subjects A4 and A9 were excluded from the analysis for two reasons. Firsrtly, the malfunctioning EEG caps, used during the experiment, made it necessary to filter out more than 50% data, including EEG data. Secondly, the amplifiers had an abnormally low common-mode rejection ratio and higher noise levels compared to other devices.
Figure 2A shows an overall increase in the F1 value during the “1–1-oT10-1” to “2–1-1–4” section, indicating an increased MF. In addition, Fig. 2C shows the individual variations in F1 values for each subject during the “1–1-oT10-1” to “2–0-open” section. Paired t-tests revealed that only A8 remained consistent (p > 0.5) with the collaborative result, while the MF levels of the other subjects differed significantly (p < 0.05) from the collaborative result. However, there was a gradual increase in MF levels during these periods. As shown in Fig. 2A and Fig. 2B, the collaborative results showed a decrease in MF during the period from “1–1-open” to “1–2-oT20”, followed by an increase from “1–3-oT15” to “2–1-1–4”. This can be attributed to the fact that the students were able to maintain their focus for 15–20 min in class, after which their MF gradually increased. Notably, F1 and F2 showed opposite trends: F1 increased over time, while F2 decreased. This finding is consistent with the conclusions of previous studies mentioned in the Methods section (Fairclough and Venables 2006; Wei et al. 2018; Eoh et al. 2005; EGELUND 1982; Lanata et al. 2015), which found that F1 was positively correlated with MF, while F2 showed a negative correlation. However, some individuals experienced a decrease in MF when learning at 2 × speed, possibly due to increased alertness. These findings will be further investigated in future studies. The observations showed a clear trend of increasing MF in several subjects.
Fig. 2 [Images not available. See PDF.]
Typical F1 and F2 results during the time. A The F1 value during a to l section, B The F2 value during a to l section, C The individual variations in F1 values during c to h section| a:1–0-Contrast-G, b:1–1-open, c:1–1-oT10–1, d:1–1-oT-2, e:1–2-oT20, f:1–3-oT15, g:1–4-oT10, h:2–0-open, i:2–1-1–1, j:2–1-1–2, k:2–1-1–3, l:2–1-1–4
As shown in Fig. 3A, there was a significant difference in F1 metrics between the “1–0-Contrast-G to 1–3-oT15” and “2–1-1–4” sections (p < 0.05). Conversely, Fig. 3B shows that there was no significant difference in F1 metrics between the sections “1–1-oT10-1 to 1–3-oT15” (p > 0.83), indicating that students maintained a consistently good MF state during this time with a certain level of concentration. Similarly, Fig. 3A shows that the F1 metrics between “1–4-oT10 to 2–1-1–1” and “1–0-Contrast-G to 1–1-open” had a similar red color, indicating no significant difference (p > 0.58). This suggests that at the “2–1-1–1” section, the MF level returned to the initial level of period 1. This observation suggests that physiological 2and psychological changes occur gradually over time, without any discrete event or external stimulus. The “2–1-1–1” sections occurred at the end of Period 2 and at the beginning of the offline teaching (approximately 40–50 min after the start of class).
Fig. 3 [Images not available. See PDF.]
Typical mutual p-value result. A Mutual T-test matrix (p-value), B P-value curve (vs section “1–1-oT10-1”) | a:1–0-Contrast-G, b:1–1-open, c:1–1-oT10-1, d:1–1-oT-2, e:1–2-oT20, f:1–3-oT15, g:1–4-oT10, h:2–0-open, i:2–1-1–1, j:2–1-1–2, k:2–1-1–3, l:2–1-1–4
Furthermore, an analysis was performed, combining the results from Fig. 2 F1 and F2 during the time, along with the mutual p-value results shown in Fig. 3. Compared to the initial sections (“1–0-contrast-G” and “1–1-open”), there was a decrease in MF (as shown in Fig. 2A and B), which peaked at section “1–3-oT15”, indicating a significant turning point. Simultaneously, the difference also approached a point near the minimum (“1–3-oT15” vs “1–0-Contrast-G”, p = 0.37), as shown in Fig. 3A. Subsequently, the MF continued to increase and returned to the baseline value (as shown in Fig. 2A and B). These results indicated a return to the baseline and a continuous deterioration in MF that cannot be reversed.
For the typical extreme value section (during “1–3-oT15”), both F1 and F2 values were examined at 1-min intervals, as shown in Fig. 2A.1 and B.1. It was evident that over these 6 min, F1 showed an increasing trend while F2 decreased, indicating an increasing MF level.
The proposed novel scale
The fitted curve was used to establish a quantification scale. Figure 4 shows how the F1 changed during the different sections, reflecting variations in the MF level. The F1 was positively correlated with MF. A second-order fitting formula was used to fit the curve of the F1 indicator data, incorporating the section as a variable. The least squares method yielded the following second-order fitting formula:
4
where s is the different section and the F1 is the MF indicator. To analyze the relationship further, looking into two specific points, “the extreme point” (s, F1) = (3.78, 1.44) and “the point equal to the initial value” (s, F1) = (8.62, 2.1763), the insights into this relationship can be gained.Fig. 4 [Images not available. See PDF.]
F1 fitting curves and boundaries | a:1–0-Contrast-G, b:1–1-open, c:1–1-oT10-1, d:1–1-oT-2, e:1–2-oT20, f:1–3-oT15, g:1–4-oT10, h:2–0-open, i:2–1-1–1, j:2–1-1–2, k:2–1-1–3, l:2–1-1–4
The results showed that at s = 3.78, the MF level reached a minimum value of 1.44. Each section corresponded to a 5-min time interval, so this point occurred at 18.9 min. Furthermore, at s = 8.62, the MF returned to the initial level at 43.1 min, which was close to the optimal class length of 45 min. The ideal lecture length appeared to exceed 15 min, possibly due to student engagement with the material presented. Combining these analyses, the common knowledge (about 15 min attention span) can be confirmed in this educational setting, and the duration of a lesson should be around 45 min.
Importantly, the group-level simultaneous data analysis in a natural class was integrated to develop a CQS with several key characteristics: continuity, quantifiability, emotion-free and high contextual relevance. The resulting quantitative intervals are shown in Figs. 2A and 4 in grey shading. To allow for dynamic individual differences in the sections, a normal interval was defined using a second-order fitting formula. The upper boundary curve (F1_up) was derived by fitting the points obtained by adding the standard deviation to F1. Conversely, the lower boundary curve (F1_down) was derived by fitting the points obtained by subtracting the standard deviation from F1. Using the least squares method, the following second-order fitting formulae (5) and (6) were derived. Formula (5) defined the upper boundary as:
5
while Formula (6) represented the lower boundary as:6
where s represented the different sections, F1_up represented the appropriate upper limit of MF, and F1_down represented the appropriate lower limit of MF.Comparison with typical work
In recent years, related research has been carried out in various aspects, including experimental design, short-term monitoring, classification methods (Wang J et al. 2023; Li J et al. 2023; Qin Y et al. 2024), materials, and hardware system design (Chen J et al. 2023; Adão Martins et al. 2021). However, there is a lack of research that combines psychological and educational assessment scales. By analyzing the related works that assess MF, concentration, distraction, or tiredness, the researches that included a long-duration task and an objective physiological indicator were selected for comparison.
The induced-fatigue tasks varied widely among studies, not only in terms of the type of fatigue studied but also in terms of application settings (as shown in Table 1, columns “Task”,“Duration”, and “Implement Type”). Induced and simulated tasks were commonly used, mainly in laboratory settings. For example, in (Choi et al. 2018), drowsiness was induced by simulating long periods of driving without a vehicle on the road, accompanied by noise. The work (Huang et al. 2018) used a 55-question quiz to manipulate MF states. The work (Li et al. 2020) used induced tasks of varying difficulty levels to induce different levels of MF. However, simulated environments cannot fully replicate the complexity and dynamics of real-world situations (Techera et al. 2018). Moreover, subjects’ effort tends to decrease in simulated environments, even in the presence of motivational factors such as monetary rewards. Furthermore, the choice and duration of an induced task lacks a scientific criterion.
Table 1. Comparisons with the state-of-the-art work
References | Task | Duration (min) | Implement type | Acquire data | Metrics | output |
|---|---|---|---|---|---|---|
Choi et al. (2018) | Induced Simulated driving | 20–120 | In lab | PPG/EDA/Motion | classification accuracy | 4 levels (normal, stressed, fatigued and/or drowsy) |
Huang et al. (2018) | Induced 55-Quiz | 54 | In lab | ECG | Questionnaire HRV | 2 levels |
Li et al. (2020) | Induced Simulated operating | 50 | In lab | Eye-tracking | Questionnaire Eye movement | 3 levels |
Ko et al. (2020) | Induced Simulated driving | 60 | In lab | EEG/EOG | Response time | 2 levels |
Torkamani-Azar et al. (2020) | Induced | Over 100 | In lab | EEG | CVS/Response time EEG Power Spectra | Prediction of Vigilance and fatigue |
Min et al. (2022) | Induced Simulated driving | Nearly 60 | In lab | EEG/ Eye-tracking | Questionnaire EEG with machine vision | 2 levels |
Yu et al. (2023) | Natural No-induced | 95 | In real class | EDA | Questionnaire EDA | 3 levels (learning efficiency) |
This work | Natural No-induced | 90 | In real class | EEG | EEG Power Spectra | Continuous Real-time assessment scales |
In contrast to previous works, our study investigated the EEG correlates of sustained fatigue and attention under real-world conditions without inducing specific tasks. A general model of sustained attention under natural human conditions was established, overcoming the limitations of previous studies. Research on dynamic cognitive performance under natural conditions is gaining increasing attention.
Conducting long-term cognitive studies is generally challenging for researchers in general (Adão Martins et al. 2021). These challenges stem from the gradual occurrence of imperceptible changes in cognitive states over time, without any distinct environmental events or external stimuli. In particular, our current research, along with another study (Yu et al. 2023), examined sustained tasks under natural conditions for a duration of 90 min and identified a gradual and continuous process of MF-related changes (see the section “The Proposed Novel Scale” for more details). To increase the ecological validity and practical significance of our findings, it is essential that future research integrates experimental data-driven findings with real-world work, meeting, or learning environments. This integration will provide a more comprehensive understanding.
Based on our research and comparisons, objective MF levels are commonly assessed using signals such as photoplethysmography (PPG), motion sensors (Motion), galvanic skin response (GSR or EDA), EEG, ECG, EOG, eye tracking, and near-infrared spectroscopy (NIRS). However, it is important to note that this article has a limitation compared to other literature, as it relies solely on unimodal EEG data. To address this limitation, future studies should use a combination of multimodal acquisition techniques (Salvi et al. 2023) under natural conditions to optimize the detection-cognitive state interval and improve the general effectiveness of suppression. In addition, the integration of MF detection and suppression with other relevant cognitive and emotional states will be explored to improve accuracy and efficiency in the future.
Regarding the metrics, the labeled data from studies conducted by these publications (Huang et al. 2018; Li et al. 2020; Torkamani-Azar et al. 2020; Min et al. 2022) were collected both before and after the tasks. This approach lacked continuous information, which hindered the understanding of MF mechanisms, although it efficiently produced a well-balanced dataset. It is important to understand that attention, MF, depression, stress, and other cognitive states are dynamic processes. An individual’s level of MF at any given moment is influenced by their previous states and can be triggered by tasks, emotions, and external stimuli. The final output consists of several cognitive levels, as shown in Table 1, column “Output”. Attention, MF, depression, stress, and other cognitive states are dynamic processes. Under this process, this work proposes a continuous, real-time, dynamic scale that is constructed through group-based activities and collaborative computation. This novel scale is then compared with other state-of-the-art scales.
Comparison with scales
Attention, alertness, and MF cognition are commonly assessed using the validated scales that allow users to record and manage within recall periods such as the past 7 days or longer, daily, timely, etc. These scales can be administered using paper-based or electronic document-based methods, both of which require retrospective recall. When comparing these scales, the factors such as the number of items, completion time, recall period, and target population were considered, as detailed in Table 2.
Table 2. Comparisons with the state-of-the-art scales
No | Scales | Items in questionnaire | Completion time (min) | Recall period | Target population |
|---|---|---|---|---|---|
#1 | Multidimensional assessment of fatigue (MAF) (Whitehead 2009) | 16 | 10 | Past 7 days | Adults with rheumatoid arthritis, multiple sclerosis, ankylosing spondylitis, and various cancers |
#2 | Chalder Fatigue Questionnaire (CFQ) (Chalder et al. 1993) | 11, 14 | 7 | Past 7 days or more | Multiple sclerosis |
#3 | Multidimensional Fatigue Inventory-20(MFI) (Smets et al. 1995) | 20 | 5–10 | Lately | People with cancer, chronic fatigue syndrome, physically tired or cognitively tired |
#4 | Epworth sleepiness scale (ESS) (Kendzerska et al. 2014) | 8 | 4 | Recently recall of drowsiness in 8 scenarios | General population with potential sleep problems, people with mental disorders |
#5 | Self-reported scale (SRS)-Self efficacy (Pintrich and De Groot 1990) | 9 | 4.5 | Timely | General adults and adolescents |
#6 | NASA Task Load Index (NASA-TLX) (Hart and Staveland 1988) | 6 | 1 | Timely | People workload assessment in aviation, engineering fields and tasks of performance |
#7 | Karolinska Sleepiness Scale (KSS) (Kaida et al. 2006) | 1 | 1 | Past 10 min | People with shift work, jetlag, driving abilities, attention |
#8 | Stanford Sleepiness Scale (SSS) (Shahid et al. 2011) | 1 | 1 | Timely | Adults over the age of 18 |
#9 | Rating Scale Mental Effort (RSME) (Ghanbary Sartang et al. 2016) | 1 | 1 | Timely | People with mental load and mental effort, such as drivers, workers, students |
#10 | the proposed CQS | 0 | 0 | Timely, real-time | General population, such as workers, students |
The number of items in the questionnaire used in traditional studies to assess MF varies depending on the target population. For example, studies of populations with conditions such as rheumatoid arthritis, multiple sclerosis, ankylosing spondylitis, various cancers, or chronic fatigue syndrome (#1, #2, #3, #4) tend to use scales with a higher number of items. In addition, the recall period for these scales is usually longer than 7 days. In studies focusing on work-related psychological disorders, scales such as #4, #5, #6, #7 and #9 have been commonly used. In the general population, scales with a more condensed number of items are commonly used to allow for timely assessment. For example, scales such as #6, #7, #8, and #9 are commonly used to measure fatigue, anxiety, stress, and workload levels in pilots, workers, students, and drivers.
In reviewing the questionnaires used in the studies listed in the “metrics” column of Table 1, variations in scale selection were observed even within studies focusing on the same population, application, or cognitive type. For example, Min et al. (Min et al. 2022) assessed driver drowsiness using the Karolinska Sleepiness Scale (KSS), while Li et al. (Li et al. 2020) used both the NASA Task Load Index (NASA-TLX) (Hart and Staveland 1988) and the Stanford Sleepiness Scale (SSS) (Shahid et al. 2011). These different choices highlight the lack of consensus on the specific cognitive dimensions to be assessed.
Considering the developmental perspective and historical progression of these studies, the understanding of assessment has evolved from primarily focusing on specific disease populations to encompassing general populations in various healthcare settings. It is evident that scales #1–#9 identify different dimensions of fatigue. However, no single measure can fully capture the intricate mechanisms and real-time process of the cognitive change. During testing, there is the possibility that individuals may inaccurately recall past experiences due to biases influenced by factors such as emotional state. In addition, subjects may tend to correlate their task performance with their scale ratings. Finally, administering the scales at the end of the task may cause subjects to forget specific task details.
Importantly, the CQS #10 is designed for a more general population. The CQS eliminates the need for items and time to complete, allowing assessments to take place outside of a laboratory setting and in real-world healthcare scenarios. Furthermore, it allows assessments to be made without memory or emotion, in real time and in a timely manner. These features address the limitations of the previously discussed scales.
Limitations and challenges
Although the comparison of the proposed CQS with the existing EEG-based MF measurement methods showed the advantages of the proposed method, it also revealed some limitations and challenges of the study. This study used its own data collected under natural conditions and did not use a public dataset, which lacks important comparisons in terms of comprehensive methods, which may reduce the generalize ability to some extent. Another limitation is the lack of standardized generally EEG data across different studies, which introduces bias and uncertainty in the comparative results. In this study, comparisons were only made in terms of “Task”, “Duration”, “Implement Type”, “Items”, “Completion time”, “Recall period”, and the experimental settings. The generalizability of the proposed methodology, the ability to measure MF levels in different tasks, subjects and environments is another limitation, for example a work task, office workers in a natural meeting; a gaming task with different skills requirements, players in a virtual reality setting. These “different tasks, subjects and environments” results would be the generalizability of the method. In addition, the generalization performance of the proposed method may be affected by other factors such as task complexity, subject variability (e.g. education) and environmental noise.
The challenge lies in the lack of a universal method for measuring MF levels, leading to inconsistencies between various studies, as different studies may use different definitions, methods, characteristics, and metrics of MF. The lack of a universally accepted EEG-based metric may further exacerbate this problem, leading to a fragmented landscape of research findings. This makes comparisons difficult and complex for general application. One possible solution is to adopt a meta-analysis, which can integrate the results of multiple studies and provide a comprehensive and objective evaluation of the different methods. These solutions can help to improve the validity and reliability of the comparison, and to advance the development of EEG-based MF measurement methods. It is also possible to generalize the performance of CQS through auto-adaptation, using the simultaneous monitoring and group-level analysis. This auto-adaptation will be explored in our future research. In addition, the integration of multimodal data sources, such as physiological and behavioral measures, may provide a more holistic view of MF and its impact on cognitive performance.
The establishment of public EEG datasets across different tasks, subjects and settings may provide a fair and consistent basis for comparison. Focusing on the development of standardized protocols for EEG data collection and universal metrics is necessary. The establishment of more ubiquitous, and granular real-time MF metrics may enhance generalization applications. The challenges highlighted in this study underscore the need for a collaborative effort within the scientific community to establish best practices for MF assessment. By working towards standardization and methodological rigor, researchers can ensure that findings are both reliable and applicable across various contexts. This will ultimately enhance the utility of CQS as a tool for assessing MF and contribute to the broader goal of understanding and mitigating the effects of cognitive fatigue.
Conclusion
Traditional measures, such as subjective questionnaires or induced-task interruption tools, lack real-time and dynamic assessment capabilities, thus failing to capture actual psychological changes. This paper proposed a CQS from the perspective of a neurocomputational measure, which was characterized by reliability, objectivity, memory- and emotion-free. Using the CQS, both the lecture duration (18.9 min) and the optimal class duration (43.1 min) were identified. These results confirmed the “common knowledge”, and highlighted the importance of group-level data-driven cognitive computing. Although this research provides valuable insights, further studies are needed to validate and generalize these findings. Particularly, it is important to conduct broader research across different age groups, educational backgrounds, and cultural environments to establish comprehensive and reliable standards for assessing attention span and MF. Our innovative findings and CQS methods represent a significant step forward in the field of educational and psychological assessment. By providing a reliable, objective, and real-time measure of MF and attention span, our work contributes to a deep understanding of cognitive dynamics in learning environments. This, in turn, provides valuable insights for optimizing educational content delivery and improving student engagement and learning outcomes.
Acknowledgements
The authors would like to express thanks to the subjects who participated in this study, and the Key-Area Research and Development Program of Guangdong Province (Project No. 2021B0909060002), the National Key R&D Program of China (Project No. 2022YFF1202303), the National Natural Science Foundation of China (Project No. 62071447) for their financial supports.
Author contributions
Yong Huang, Yijun Wang, Lirong Zheng: study conception and design; Yong Huang: data collection and visualization; Yijun Wang, Xiaorong Gao: experiment validation; Yuxiang Huan, Yijun Wang, Zhuo Zou and Lirong Zheng: writing, review and editing; Yong Huang, Yuxiang Huan: analysis and interpretation of results.
Data availability
The raw data supporting the conclusions of this article will be made available on reasonable request.
Declarations
Conflict of interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
Adão Martins, NR; Annaheim, S; Spengler, CM; Rossi, RM. Fatigue monitoring through wearables: a state-of-the-art review. Front Physiol; 2021; 12, pp. 790292-790292. [DOI: https://dx.doi.org/10.3389/fphys.2021.790292] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34975541][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8715033]
Ali S, Iqbal KF, Avaz Y, Saiid M (2022) A review on different approaches for assessing student attentiveness in classroom using behavioural elements. IEEE 2022 2nd International Conference on Artificial Intelligence (ICAI). pp 152–158. https://doi.org/10.1109/ICAI55435.2022.9773418
Argyriou, P; Benamar, K; Nikolajeva, M. What to blend? Exploring the relationship between student engagement and academic achievement via a blended learning approach. Psychol Learn Teach; 2022; 21, pp. 126-137. [DOI: https://dx.doi.org/10.1177/14757257221091512]
Atwa, H; Shehata, MH; Al-Ansari, A et al. Online, face-to-face, or blended learning? Faculty and medical students’ perceptions during the COVID-19 pandemic: a mixed-method study. Front Med; 2022; 9, pp. 791352-791352. [DOI: https://dx.doi.org/10.3389/fmed.2022.791352]
Balandong, RP; Ahmad, RF; Mohamad Saad, MN; Malik, AS. A review on EEG-based automatic sleepiness detection systems for driver. IEEE Access; 2018; 6, pp. 22908-22919. [DOI: https://dx.doi.org/10.1109/access.2018.2811723]
Borghini, G; Astolfi, L; Vecchiato, G et al. Measuring neurophysiological signals in aircraft pilots and car drivers for the assessment of mental workload, fatigue and drowsiness. Neurosci Biobehav Rev; 2014; 44, pp. 58-75. [DOI: https://dx.doi.org/10.1016/j.neubiorev.2012.10.003] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23116991]
Bradbury, NA. Attention span during lectures: 8 seconds, 10 minutes, or more?. Adv Physiol Educ; 2016; 40, pp. 509-513. [DOI: https://dx.doi.org/10.1152/advan.00109.2016] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28145268]
Chalder, T; Berelowitz, G; Pawlikowska, T et al. Development of a fatigue scale. J Psychosom Res; 1993; 37, pp. 147-153.
Chen, J; Xu, B; Zhang, D. Inter-brain coupling analysis reveals learning-related attention of primary school students. Educ Tech Res Dev; 2023; [DOI: https://dx.doi.org/10.1007/s11423-023-10311-3]
Choi, M; Koo, G; Seo, M; Kim, SW. Wearable device-based system to monitor a driver’s stress, fatigue, and drowsiness. IEEE Trans Instrum Meas; 2018; 67, pp. 634-645. [DOI: https://dx.doi.org/10.1109/tim.2017.2779329]
da Silveira, TLT; Kozakevicius, AJ; Rodrigues, CR. Automated drowsiness detection through wavelet packet analysis of a single EEG channel. Expert Syst Appl; 2016; 55, pp. 559-565. [DOI: https://dx.doi.org/10.1016/j.eswa.2016.02.041]
Egelund, N. Spectral analysis of heart rate variability as an indicator of driver fatigue. Ergonomics; 1982; 25, pp. 663-672.
Eoh, HJ; Chung, MK; Kim, S-H. Electroencephalographic study of drowsiness in simulated driving with sleep deprivation. Int J Ind Ergon; 2005; 35, pp. 307-320. [DOI: https://dx.doi.org/10.1016/j.ergon.2004.09.006]
Eronen, MI; Bringmann, LF. The theory crisis in psychology: how to move forward. Perspect Psychol Sci J Assoc Psychol Sci; 2021; 16, pp. 779-788. [DOI: https://dx.doi.org/10.1177/1745691620970586]
Fairclough, SH; Venables, L. Prediction of subjective states from psychophysiology: a multivariate approach. Biol Psychol; 2006; 71, pp. 100-110. [DOI: https://dx.doi.org/10.1016/j.biopsycho.2005.03.007] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15978715]
Fairclough, SH; Venables, L; Tattersall, A. The influence of task demand and learning on the psychophysiological response. Int J Psychophysiol; 2005; 56, pp. 171-184. [DOI: https://dx.doi.org/10.1016/j.ijpsycho.2004.11.003] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15804451]
Ghanbary Sartang, A; Ashnagar, M; Habibi, E; Sadeghi, S. Evaluation of rating scale mental effort (RSME) effectiveness for mental workload assessment in nurses. J Occup Health Epidemiol; 2016; 5,
Gopi, Y; Madan, CR. Subjective memory measures: Metamemory questionnaires currently in use. Q J Exp Psychol; 2023; [DOI: https://dx.doi.org/10.1177/17470218231183855]
Habay, J; Van Cutsem, J; Verschueren, J et al. Mental fatigue and sport-specific psychomotor performance: a systematic review. Sports Med; 2021; 51, pp. 1527-1548. [DOI: https://dx.doi.org/10.1007/s40279-021-01429-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33710524]
Hamann, A; Carstengerdes, N. Assessing the development of mental fatigue during simulated flights with concurrent EEG-fNIRS measurement. Sci Rep; 2023; 13, 4738.
Hart, SG; Staveland, LE. Development of NASA-TLX (Task load index) results of empirical and theoretical research. Adv Psychol; 1988; 52, pp. 139-183. [DOI: https://dx.doi.org/10.1016/s0166-4115(08)62386-9]
Hartley, J; Davies, IK. Note-taking: a critical review. Program Learn Educ Technol; 1978; 15, pp. 207-224. [DOI: https://dx.doi.org/10.1080/0033039780150305]
Hu, B; Shen, J; Zhu, L et al. Fundamentals of computational psychophysiology: theory and methodology. IEEE Trans Comput Soc Syst; 2022; 9, pp. 349-355. [DOI: https://dx.doi.org/10.1109/tcss.2022.3157522]
Huang, S; Li, J; Zhang, P; Zhang, W. Detection of mental fatigue state with wearable ECG devices. Int J Med Inf; 2018; 119, pp. 39-46. [DOI: https://dx.doi.org/10.1016/j.ijmedinf.2018.08.010]
Huang, Y; Huan, Y; Zou, Z et al. A wearable group-synchronized EEG system for multi-subject brain-computer interfaces. Front Neurosci; 2023; 17, pp. 1176344-1176344. [DOI: https://dx.doi.org/10.3389/fnins.2023.1176344] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37539380][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10396297]
Kaida, K; Takahashi, M; Åkerstedt, T et al. Validation of the Karolinska sleepiness scale against performance and EEG variables. Clin Neurophysiol; 2006; 117, pp. 1574-1581. [DOI: https://dx.doi.org/10.1016/j.clinph.2006.03.011] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16679057]
Kendzerska, TB; Smith, PM; Brignardello-Petersen, R et al. Evaluation of the measurement properties of the Epworth sleepiness scale: a systematic review. Sleep Med Rev; 2014; 18, pp. 321-331. [DOI: https://dx.doi.org/10.1016/j.smrv.2013.08.002] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24135493]
Kikuchi, H; Odagiri, Y; Ohya, Y et al. Association of overtime work hours with various stress responses in 59,021 Japanese workers: retrospective cross-sectional study. PLoS ONE; 2020; 15, pp. e0229506-e0229506.
Kleinsasser, RC; McKeachie, WJ. Teaching tips: Strategies, research, and theory for college and university teachers. Mod Lang J; 1994; 78, 545. [DOI: https://dx.doi.org/10.2307/328598]
Knox, L; Berzenski, S; Drew, S. Measuring zoom fatigue in college students: development and validation of the meeting fatigue scale for videoconferencing (MFS-V) and the meeting fatigue scale for in-person (MFS-I). Media Psychol; 2023; 26, pp. 680-712. [DOI: https://dx.doi.org/10.1080/15213269.2023.2204529]
Ko, L-W; Komarov, O; Hairston, WD et al. Sustained attention in real classroom settings: an EEG study. Front Hum Neurosci; 2017; 11, pp. 388-388. [DOI: https://dx.doi.org/10.3389/fnhum.2017.00388] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28824396][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5534477]
Ko, L-W; Komarov, O; Lai, W-K et al. Eyeblink recognition improves fatigue prediction from single-channel forehead EEG in a realistic sustained attention task. J Neural Eng; 2020; 17, 036015. [DOI: https://dx.doi.org/10.1088/1741-2552/ab909f] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32375139]
Lanata, A; Valenza, G; Greco, A et al. How the autonomic nervous system and driving style change with incremental stressing conditions during simulated driving. IEEE Trans Intell Transp Syst; 2015; 16, pp. 1505-1517. [DOI: https://dx.doi.org/10.1109/tits.2014.2365681]
Lerdal, A. Maggino, F. Fatigue severity scale. Encyclopedia of quality of life and well-being research; 2021; Cham, Springer: [DOI: https://dx.doi.org/10.1007/978-3-319-69909-7_1018-2]
Li, J; Li, H; Umer, W et al. Identification and classification of construction equipment operators’ mental fatigue using wearable eye-tracking technology. Autom Constr; 2020; 109, 103000. [DOI: https://dx.doi.org/10.1016/j.autcon.2019.103000]
Li, J; Li, Y; Du, M. Comparative study of EEG motor imagery classification based on DSCNN and ELM. Biomed Signal Process Control; 2023; 84, 104750. [DOI: https://dx.doi.org/10.1016/j.bspc.2023.104750]
Maddox, H; Hoole, E. Performance decrement in the lecture. Educ Rev; 1975; 28, pp. 17-30. [DOI: https://dx.doi.org/10.1080/0013191750280102]
Masri, G; Al-Shargie, F; Tariq, U et al. Mental stress assessment in the workplace: a review. IEEE Trans Affect Comput; 2023; [DOI: https://dx.doi.org/10.1109/TAFFC.2023.3312762]
Min, J; Cai, M; Gou, C et al. Fusion of forehead EEG with machine vision for real-time fatigue detection in an automatic processing pipeline. Neural Comput Appl; 2022; [DOI: https://dx.doi.org/10.1007/s00521-022-07466-0] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36467631][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9684777]
Nagasaki, K; Nishizaki, Y; Shinozaki, T et al. Association between mental health and duty hours of postgraduate residents in Japan: a nationwide cross-sectional study. Sci Rep; 2022; 12, pp. 10626-10626.
Nesher Shoshan, H; Wehrt, W. Understanding “Zoom fatigue”: a mixed-method approach. Appl Psychol; 2021; 71, pp. 827-852. [DOI: https://dx.doi.org/10.1111/apps.12360]
Pintrich, PR; De Groot, EV. Motivational and self-regulated learning components of classroom academic performance. J Educ Psychol; 1990; 82, pp. 33-40. [DOI: https://dx.doi.org/10.1037/0022-0663.82.1.33]
Qin, Y; Yang, B; Ke, S et al. M-FANet: Multi-feature attention convolutional neural network for motor imagery decoding. IEEE Trans Neural Syst Rehabil Eng; 2024; 32, pp. 401-411. [DOI: https://dx.doi.org/10.1109/TNSRE.2024.3351863] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38194394]
Salvi, M; Loh, HW; Seoni, S et al. Multi-modality approaches for medical support systems: a systematic review of the last decade. Inf Fusion; 2023; [DOI: https://dx.doi.org/10.1016/j.inffus.2023.102134]
Schwartz, AC; Cotes, RO; Kim, J et al. Bite-sized teaching: engaging the modern learner in psychiatry. Acad Psychiatry; 2019; 43, pp. 315-318. [DOI: https://dx.doi.org/10.1007/s40596-018-1014-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30607895]
Shahid, A; Wilkinson, K; Marcu, S; Shapiro, CM. Shahid, A; Wilkinson, K; Marcu, S; Shapiro, C. Stanford sleepiness scale (SSS). STOP, THAT and One Hundred Other Sleep Scales; 2011; New York, Springer: [DOI: https://dx.doi.org/10.1007/978-1-4419-9893-4_91]
Smets, EMA; Garssen, B; Bonke, B; De Haes, JCJM. The multidimensional fatigue inventory (MFI) psychometric qualities of an instrument to assess fatigue. J Psychosom Res; 1995; 39, pp. 315-325.
Sun, H; Soh, KG; Roslan, S; Wazir, MRWN; Soh, KL. Does mental fatigue affect performance among individual sport athletes? A systematic review. PLoS ONE; 2021; 16, pp. e0258307-e0258307.
Szpunar, KK; Moulton, ST; Schacter, DL. Mind wandering and education: from the classroom to online learning. Front Psychol; 2013; 4, 495. [DOI: https://dx.doi.org/10.3389/fpsyg.2013.00495] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23914183][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3730052]
Techera, U; Hallowell, M; Littlejohn, R; Rajendran, S. Measuring and predicting fatigue in construction: empirical field study. J Constr Eng Manag; 2018; 144,
Tian S, Wang Y, Dong G, et al (2018) Mental fatigue estimation using EEG in a vigilance task and resting states. 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. pp 1980–1983. https://doi.org/10.1109/EMBC.2018.8512666
Torkamani-Azar, M; Kanik, SD; Aydin, S; Cetin, M. Prediction of reaction time and vigilance variability from spatio-spectral features of resting-state EEG in a long sustained attention task. IEEE J Biomed Health Inform; 2020; 24, pp. 2550-2558. [DOI: https://dx.doi.org/10.1109/jbhi.2020.2980056] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32167917]
Vekaria, PC; Peverly, ST. Lecture note-taking in postsecondary students with attention-deficit/hyperactivity disorder. Read Writ; 2018; 31, pp. 1551-1573. [DOI: https://dx.doi.org/10.1007/s11145-018-9849-2]
Wang, J; Cheng, S; Tian, J et al. A 2D CNN-LSTM hybrid algorithm using time series segments of EEG data for motor imagery classification. Biomed Signal Process Control; 2023; 83, 104627. [DOI: https://dx.doi.org/10.1016/j.bspc.2023.104627]
Wei, C-S; Wang, Y-T; Lin, C-T; Jung, T-P. Toward drowsiness detection using non-hair-bearing EEG-based brain-computer interfaces. IEEE Trans Neural Syst Rehabil Eng; 2018; 26, pp. 400-406. [DOI: https://dx.doi.org/10.1109/tnsre.2018.2790359] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29432111]
Whitehead, L. The measurement of fatigue in chronic illness: a systematic review of unidimensional and multidimensional fatigue measures. J Pain Symptom Manage; 2009; 37, pp. 107-128. [DOI: https://dx.doi.org/10.1016/j.jpainsymman.2007.08.019] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/19111779]
Wilson, K; Korn, JH. Attention during lectures: beyond ten minutes. Teach Psychol; 2007; 34, pp. 85-89. [DOI: https://dx.doi.org/10.1177/009862830703400202]
Wingelaar-Jagt, YQ; Wingelaar, TT; Riedel, WJ; Ramaekers, JG. Fatigue in aviation: safety risks, preventive strategies and pharmacological interventions. Front Physiol; 2021; 12, pp. 712628-712628. [DOI: https://dx.doi.org/10.3389/fphys.2021.712628] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34552504][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8451537]
Yu, H; Xu, M; Xiao, X et al. Detection of dynamic changes of electrodermal activity to predict the classroom performance of college students. Cogn Neurodyn; 2023; 18, pp. 173-184. [DOI: https://dx.doi.org/10.1007/s11571-023-09930-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38406194]
Zhang, X; Lu, D; Pan, J et al. Fatigue detection with covariance manifolds of electroencephalography in transportation industry. IEEE Trans Ind Inform; 2021; 17, pp. 3497-3507. [DOI: https://dx.doi.org/10.1109/tii.2020.3020694]
Zhang, X; Pan, J; Shen, J et al. Fusing of electroencephalogram and eye movement with group sparse canonical correlation analysis for anxiety detection. IEEE Trans Affect Comput; 2022; 13, pp. 958-971. [DOI: https://dx.doi.org/10.1109/taffc.2020.2981440]
Zhao, C; Zhao, M; Liu, J; Zheng, C. Electroencephalogram and electrocardiograph assessment of mental fatigue in a driving simulator. Accid Anal Prev; 2012; 45, pp. 83-90. [DOI: https://dx.doi.org/10.1016/j.aap.2011.11.019] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/22269488]
Zorowitz, S; Niv, Y. Improving the reliability of cognitive task measures: a narrative review. Biol Psychiatry: Cogn Neurosci Neuroimaging; 2023; 8, pp. 789-797. [DOI: https://dx.doi.org/10.1016/j.bpsc.2023.02.004] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36842498]
Copyright Springer Nature B.V. Dec 2024