Functional network changes associated with sleep deprivation and fatigue during simulated driving: Validation using blood biomarkers
Introduction
In recent years, considerable research has been carried out to study fatigue and sleep deprivation in human drivers using several channels of information. Some of these, such as ocular measures, electroencephalogram (EEG), electrocardiogram (ECG), electromyogram (EMG), and skin conductance, turn out to be somewhat informative about the state of fatigue or drowsiness in an individual. Fatigue occurs gradually as a person starts working. Especially, with a monotonous type of work, the person becomes frustrated after a period of time and loses motivation. This is followed by a feeling of tiredness and, eventually, drowsiness before sleep overcomes the individual. Generally, it is difficult to assess the exact state of fatigue of an individual from the above-mentioned channels of information. However, currently, the EEG has been considered to be the most reliable indicator of the state of fatigue in humans (Lal and Craig, 2001).
Several methods have been reported in the literature for detecting fatigue and sleepiness using EEG signals. Most of them involve calculation of features such as energy and entropy in different frequency bands (Siemionow et al., 2004, Papadelis et al., 2006). These parameter-based approaches are useful to characterise local variations. However, they fail to characterise the interactive dynamics in the brain. In this work, a network-based approach has been adopted to study the transition of brain dynamics as a person passes from an alert to a fatigued state (Stam and Reijneveld, 2007, Strogatz, 2001). The interaction between the signals from different electrodes has been quantified using a non-linear measure known as the synchronisation likelihood (SL). This method is suitable for studying non-stationary signals such as EEG (Stam and van Dijk, 2002). Further, the network characteristic has been quantified using a number of standard parameters such as degree of connectivity, cluster coefficient and mean path length. (Stam and Reijneveld, 2007, Boccaletti et al., 2006).
EEG oscillations are the basic means of communication between cortical cell assemblies. The nature of oscillations in different frequency bands is related to various tasks and performances. For example, the frequency of the alpha rhythm is related to memory performance (Klimesch, 1996). Moreover, different types of synchronisation between various areas of brain have been observed during different types of activities. For example, a significant higher level of de-synchronisation is reported during complex and difficult tasks than during less complex and easy tasks. During mental activity, different neuronal networks begin to oscillate at different frequencies, whereas synchronous oscillations of large cell assemblies have been observed during the resting state or functional inhibition. This is true for both motor and cognitive types of tasks (Klimesch, 1996). From the above-mentioned facts, it is envisaged that the synchronisation of EEG signals from various locations might provide an insight into the gradual development of fatigue as well as sleepiness.
The EEG signal, being highly non-stationary, is very difficult to be characterised completely either in time or in frequency domain. In recent times, wavelet transform has been extensively used in EEG signal analysis (Al-Nashash et al., 2002, Geva and Kerem, 1998, Yamaguchi, 2003). This technique provides a multi-resolution time-scale representation of the signal (Mallat S, 1999, Daubechies, 1990). In this article, we have applied the discrete wavelet transform (DWT) to decompose the EEG signal into various bands and then apply SL for generating the spatial network topology among the various EEG channels.
Various blood biochemical parameters, particularly the neurotransmitters and their metabolic products, blood glucose, serum urea and creatinine, represent either the inducers or the outcome of events leading to fatigue. These parameters can reliably reflect the gradual genesis of central as well as peripheral fatigue. Researchers have already established marked alterations in glucose metabolism, in some situations resembling patients with type-2 diabetes, during sleep deprivation (Gangwisch et al., 2007).
In this article, we first pass the signal through a band-pass filter followed by DWT decomposition. Second, we quantify the interdependency of signals from different electrodes using SL measures and construct the network. Third, we evaluate standard parameters to characterise the networks pertaining to different stages of fatigue across all subjects. Finally, we perform a correlation study between network parameters and blood biochemical parameters.
The article has been organised as follows:
Section 2 describes the experimental method and data collection. In Section 3, the methodology of analysis has been described, which includes signal pre-processing, signal decomposition using DWT, network formation using SL, network-characterising parameters and biochemical analysis of blood parameters. Section 4 describes the results along with discussions. The correlation study has been presented in Section 5.
Section snippets
Subjects
A total of 12 healthy male subjects in the age group of 20–35 years have been chosen for the experiment. All the subjects were reported to have no sleep-related disorders. They had normal visual acuity. Their fitness and health were checked thoroughly by a medical practitioner before their selection as well as during the experiment. The selected subjects were advised to maintain a prescribed routine during 48 h prior to the experiment. They were asked to refrain from consuming any type of
Methodology of analysis
Our process involves filtering, decomposition into different bands by DWT, network formation using SL, computation of network parameters and biochemical analysis of blood parameters. The detailed methodology is explained in the following sections.
Results
The analysis has been carried out on 3-min EEG records (during the computer game) of 12 subjects at 11 stages (stages 2–12). The temporal values of SL between a pair of electrodes have been evaluated for each frequency band. The time-averaged SL values have been used to construct the SL matrix. The mean SL matrices of all subjects in various bands at all stages of the experiment are shown in Fig. 1(a). The corresponding adjacency matrices and the networks, as shown in Fig. 1b, Fig. 1c,
Correlation Study
A statistical correlation analysis (Cohen et al., 2003) has been carried out to correlate the trend of blood parameters with the trend of EEG network parameters at five stages (stages 1, 3, 6, 9 and 11 corresponding to the time of blood sample collection) of the experiment. The correlation coefficient has been computed between mean (mean of all subjects) blood parameter values and mean EEG parameter values at five stages (stages 1, 3, 6, 9 and 11). The correlation result is shown in Table 1. It
Discussion
The rhythmic neural oscillations play a crucial role in the transfer of information between different regions of the brain. Such oscillations from different regions show global synchronisation (type 1 synchronisation), local synchronisation (type 2 synchronisation) or de-synchronisation (Klimesch, 1996) depending on the nature of physical and mental tasks. It has been observed that cognitive tasks or complex information processing requires rapidly changing and widely distributed neural
Conclusions
The major finding of the above study is the synchronisation of specific bands of the EEG signals from different cortical areas as a result of sleep deprivation and fatigue with simultaneous validation using blood biochemical parameters. The results from network analysis suggest an increase in the degree of connectivity, clustering coefficient, small-world metric and a decrease in characteristic path length in some frequency bands of the signals at increased levels of fatigue and sleepiness. The
Acknowledgement
We are grateful to the reviewers for helping us to improve the article quality. We acknowledge the financial support from Industrial Working Group, Department of Information Technology, Government of India, through the sponsored project in ‘Development of Real Time Algorithms for Analysis of Fatigue in Human Drivers’. We also thank all the subjects for their co-operation and the members of our research team for their valuable suggestions and support in the design of experiment and collection of
References (27)
- et al.
Complex networks: Structure and dynamics
Phys Rep
(2006) - et al.
Effects of mental fatigue on attention: an ERP study
Cogn Brain Res
(2005) Memory processes, brain oscillations and EEG synchronization
Int J Psychophysiol
(1996)- et al.
A critical review of the psychophysiology of driver’s fatigue
Biol Psychol
(2001) - et al.
The influence of mental fatigue and motivation on neural network dynamics; an EEG coherence study
Brain Res
(2009) - et al.
Synchronization Likelihood with explicit time-frequency priors
NeuroImage
(2006) - et al.
Altered central nervous system signal during motor performance in chronic fatigue syndrome
Clin Neurophysiol
(2004) - et al.
Organization, development and function of complex brain networks
Trends Cogn Sci
(2004) Functional connectivity patterns of human magnetoencephalographic recordings: a small world network
Neurosci Lett
(2004)- et al.
Synchronization Likelihood: an unbiased measure of generalized synchronization of multivariate data set
Physica D
(2002)
A wavelet chaos methodology for analysis of EEGs and EEG subbands to detect seizure and epilepsy
IEEE Trans Bio-Med Eng
Understanding cognition through large scale cortical network
Curr Dir Psychol Sci
Cited by (39)
Driver fatigue detection based on prefrontal EEG using multi-entropy measures and hybrid model
2021, Biomedical Signal Processing and ControlNeural Mechanisms of Mental Fatigue Revisited: New Insights from the Brain Connectome
2019, EngineeringCitation Excerpt :Given the paucity of studies on the neural mechanisms of fatigue recovery, further investigations with careful experiment design considering various break manipulation factors are recommended in order to confirm our findings. Connectivity studies have generally found evidence of reduced global integration (higher path length and/or lower global efficiency [39,41,71,74,95,96,108], although see Ref. [88] for an exception) and increased local segregation (clustering coefficient and/or local efficiency [41,71,74,88,95,108]) in the fatigue state (Table 2). According to the global neuronal workspace theory [40], a globally integrated network topology is needed to support high task demands.
Wavelet multiresolution complex network for decoding brain fatigued behavior from P300 signals
2018, Physica A: Statistical Mechanics and its ApplicationsCitation Excerpt :In particular, a functional brain network is constructed by regarding brain regions (electrodes of scalp EEG) as nodes and the interrelations between each pair of nodes as edges. In [41], functional brain network changes associated with sleep deprivation and fatigue are explored. In [42], driving fatigue is estimated by brain network analysis.
An adaptive optimal-Kernel time-frequency representation-based complex network method for characterizing fatigued behavior using the SSVEP-based BCI system
2018, Knowledge-Based SystemsCitation Excerpt :In [32], a Granger-Causality-based network is used to analyze the mental state of fatigue driving. In [33], the changes of brain functional network between sleep deprivation and fatigue driving are studied. A review of complex network analysis of time series refers to Ref. [34].
A systematic review for the fatigue driving behavior recognition method
2024, Journal of Intelligent and Fuzzy Systems