Comparing combinations of EEG activity in train drivers during monotonous driving

Abstract

This study investigated the changes in electroencephalography (EEG) activity in train drivers during a monotonous train-driving session. Four combinations of EEG activities were also compared to investigate the difference in performance of these equations. The four equations tested were equation 1 (θ/β), equation 2 (θ/(α + β)), equation 3 ((θ + α)/β), and equation 4 ((θ + α)/(α + β)). A total of fifty male train drivers were recruited to perform a 30-min monotonous train-driving task while 2-channels of EEG (frontal and temporal) were recorded. At the frontal site, significant differences were found for theta (p = 0.045) and alpha (0.0001) activities, and at the temporal site, significant differences were found for delta (p = 0.007) and theta (0.01) activities. For the average of frontal and temporal site activities, significant differences were found for delta (p = 0.004), theta (p = 0.001), and beta (p = 0.048). Significant difference were found for temporal site for equation 1 (θ/β) (p = 0.04), and equation 4 ((θ + α)/(α + β)) (p = 0.02), and for the average of frontal and temporal site activities, significant differences were found for all four equations (equation 1 (p = 0.001), equation 2 (p = 0.006), equation 3 (p = 0.04), and equation 4 (p = 0.002)). These findings can be utilised as a potential fatigue indicator.

Introduction

Train accidents can be caused by several reasons, including poor track condition (Bruzelius & Mba, 2004), faulty train machinery or wheels (Esslinger, Kieselbach, Koller, & Weisse, 2004), and human errors (Edkins and Pollock, 1997, Wilde and Stinson, 1983). However, nearly 75% of all train accidents can be attributed to human error, which is mostly caused by fatigued train crew members (Edkins and Pollock, 1997, Wilde and Stinson, 1983). There is strong evidence supporting a relationship between railway accidents and locomotive crew fatigue, such as time on duty, work–sleep–rest cycles and shift work (Wilde & Stinson, 1983). Studies have shown that a number of train drivers were fatigued while driving the train, especially when driving at night-time (Austin and Drummond, 1986, Buck and Lamonde, 1993, Rajaratnam and Jones, 2004, Smiley, 1990, Torsvall and Åkerstedt, 1987).

Härmä, Sallinen, Ranta, Mutanen, and Müller (2002) have shown that train drivers suffered sleepiness and fatigue with irregular shift schedules. Austin and Drummond (1986) found that about 25% of train drivers dozed off while driving or waiting at the station. Fatigue and sleepiness are more prevalent in night shift-workers (Åkerstedt, 1988, Costa, 1997, Härmä et al., 2002). The tendency of falling asleep has been shown to be at least 6 times higher in the night shift, and at least double in the morning shift, when compared to the traditional day shift (Härmä et al., 2002).

The irregular train driver shift may be managed through the use of fatigue management software, such as the Fatigue Audit InterDyne (FAID) software, which calculates the approximate fatigue level of train drivers and manages the train drivers’ rosters to minimise the fatigue level that drivers experience while on the job (Fletcher & Dawson, 1997). However, the software cannot account for drivers’ activities outside the work hours, and hence, the FAID software’s prediction of the drivers’ fatigue levels when working may not be accurate (Australian Transport Safety Bureau (ATSB), 2007). Several train mishaps that occurred in Australia have been associated with fatigue, such as the Beresfield (Australian Transport Safety Bureau (ATSB), 1998), Epping (Australian Transport Safety Bureau (ATSB), 2003), Waterfall (McInerney, 2005), and Benalla (Australian Transport Safety Bureau (ATSB), 2007) train accidents. After such train accidents, FAID score was used in the post-accident investigation to identify pre-accident fatigue levels in the train drivers’ and train guards’. The drivers involved in the Epping and Benalla accidents had FAID scores well below 80, which was the acceptable FAID score (ATSB, 2003, ATSB, 2007). However, fatigue was still believed to be the cause of the accidents (ATSB, 2003, ATSB, 2007). There are also reports of fatigue-related train accidents from other countries, such as USA (NTSB, 2005, NTSB, 2006), and United Kingdom (RAIB, 2007a, RAIB, 2007b).

Various researches into developing fatigue countermeasure devices exist. Some studies have focused on the rate of eye blink and the percentage of eye closure as an indicator of fatigue (Russo et al., 2003, Stern, 1994, Wierwille et al., 1994). Others have proposed the extraction of facial features in order to detect fatigue (Eriksson and Papanikolopoulos, 2001, Gu et al., 2002, Singh and Papanikolopoulos, 1999), while others have combined facial feature extraction techniques and measurement of eye closure and the rate of eye blinks to detect fatigue (Longhurst, 2002). However, fatigue can occur in absence of eye closures and any visible physical signs of fatigue (Horne and Reyner, 1999, Williamson and Chamberlain, 2005).

Electroencephalography (EEG) has been proposed as an accurate indicator of fatigue (Artaud et al., 1994, Lal and Craig, 2002). When one is alert, the brain activities show more fast wave (alpha (8–13 Hertz (Hz)) and beta (13–35 Hz)) activities (Kiymik, Akin, & Subasi, 2004), while a fatigued individual shows more slow wave (delta (0–4 Hz) and theta (4–8 Hz)) activities (Santamaria & Chiappa, 1987). An increase of theta activity and a decrease in alpha and beta activities have been associated with a deterioration of driving performance (Davies, 1965). Others have combined the EEG frequency bands to form an equation, such as (θ + α)/β and β/α, that may be used in a fatigue detection technique (Brookhuis and Waard, 1993, Eoh et al., 2005). Eoh et al. (2005) believed that the equation (θ + α)/β was a reliable fatigue indicator, since it combined the theta and alpha power to detect changes in alertness level, and Jap, Lal, Fischer, and Bekiaris (2009) have shown that equation (θ + α)/β is a reliable indicator of fatigue in non-professional drivers.

The aim of the present study was to investigate the changes in EEG activity (delta, theta, alpha, and beta) in train drivers during a monotonous train-driving session, and to investigate the performance difference in several combinations of EEG activity that might be used as a potential fatigue indicator.