Anda di halaman 1dari 20

23/10/2012

Medicine, Nursing and Health Sciences

Fatigue Detection Technologies


Clare Anderson, Ph.D. 1,2,3
1Sleep 2

& Chronobiology Research group, Monash University Brigham and womens Hospital, Boston, USA. 3 Harvard Medical School, Boston, USA

Fatigue: An Overview
1. Acute Sleep Deprivation 2. Chronic Sleep Deprivation 3. Time of Day 4. Sleep inertia
Performance degrades due to time awake

Performance degrades cumulatively with each day that sleep is inadequate Performance is markedly worse during nighttime hours Performance is worse upon awakening.

Fatigue Detection Technologies

15th September 2012

23/10/2012

Fatigue Management - approaches


Maintaining alertness: an integrated system

Prediction
Screening techniques Biomathematical models Fitness for duty testing

Prevention
Comprehensive Education Programs Fitness for duty testing

Detection
Fatigue Detection Technologies Employee or Peer Feedback

Intervention
Roster scheduling Intervention schemes Countermeasures

Fatigue Detection Technologies

15th September 2012

Importance of fatigue detection technologies


Brown (1997) three main reasons: 1. Fatigue is a persistent occupational hazard 2. Pressure to complete task, despite growing levels of fatigue 3. Detrimental impact of fatigue on performance PROs
Important part of fatigue management Subjective reports are less reliable, and prone to error Act as: Detection, Deterrent, and Intervention

CONs
Reliability/validity is largely unknown in operational conditions Could be abused (Brown, 1997)
a technology that cab potentially enhance safety and save lives should not be prejudged based on speculations about users ethics, whether realistic or exaggerated Dinges & Maillis, 1998. p.210
Fatigue Detection Technologies 15th September 2012 4

23/10/2012

Development of Technologies
Four major steps of technology development:

1) Discovery and identification of biomarkers (outcomes)

2) Technology development

3) Scientific evaluation of the technology

4) Operational/ecological validation of the technology

Fatigue Detection Technologies

15th September 2012

Scientific validation of technologies


Problems and Requirements PROBLEMS
Dinges & Mallis (1998)

What to validate the technology against? Adequacy of the validation data threshold setting. REQUIREMENTS
Validity Concurrent Validity Predictive Validity Reliability Generalisability Sensitivity Specificity
Does it measure what it is supposed to measure? Does the technology convey the same level of fatigue as another known measure at the same point in time? Does the technology predict a level of fatigue at a future point in time? Does the technology show consistent outcomes under similar conditions Does the technology measure the outcome in every individual How often will the technology miss detecting a fatigue event? How often will the technology give an alarm that is false?
Fatigue Detection Technologies 15th September 2012 6

23/10/2012

Fatigue Detection Technologies:


Options for industry 22 fatigue detection technologies subjected to expert review under 26 categories
Focus of technology; System capabilities; primary sensory technology; primary measures; primary system characteristics; system integration requirements; fatigue countermeasures; environmental requirements; data evaluation/recording; validation and system accuracy; operator acceptance; data integration/calibration; technology readiness.

Top performing technologies subjected to evaluation

Fatigue Detection Technologies

15th September 2012

Operator Fatigue Detection Technology Review


Caterpillar 2008

Fatigue Technology HaulCheck PASS Fatigue Management System Safe Trac NOV Alert Driver Fatigue Monitor (CoPilot) Driver State Monitor Voice Commander System Lane Departure EyeCheck Vision/Radar Sensor EDVTCS OSPAT ASTIDTM SleepWatch Fatigue Warning System Sleep Control Helmet System Facelab OptalertTM Antisleep Insight NapZapper

Company AcuMine Advanced Safety Concepts ARRB Transport Research Assistware Technologies Atlas Research ltd Attention Technologies Delphi Corporation International Mining Technologies Iteris Inc MCJ Mobileye NV Neurocom Ospat Pty Pernix/FMI Precision Control Design Inc Muirhead/Remote Control tech. Security eEectronic Systems Seeing Machines Sleep Diagnostics pty Smart eye SMI Welkin

Parameter of Interest Lane Deviation Head Nodding detection Reaction Time Lane deviation Muscle Tone Analysis Eye Blink Detection Eye Blink Detection Reaction Time Lane Deviation Fitness for Duty Lane Deviation Skin Conductance Fitness for Duty Steering Movement Activity Monitor Reaction Time Head Nodding Eye Movements Eye Movement Eye and Head Movements Eye and Head Movements Head Nodding

Based on table 1. Caterpillar (2008)


15th September 2012 8

Fatigue Detection Technologies

23/10/2012

Outcome Parameters: Scientific Validity


Muscle tone (n=1) Skin Conductance (n=1) Activity Monitor (n=1) Fitness for Duty (n=2) Head Nodding (n=3) Reaction Time (n=3) Lane Deviation (n=5) Eye Movements (n=6)
Taken from www.smartmotorist.com 06/09/2012

Fatigue Detection Technologies

15th September 2012

Operator Fatigue Detection Technology Review


Caterpillar 2008 ASTiD Seeing Machines Optalert SmartCap

Fatigue Technology ASTIDTM Facelab HaulCheck OptalertTM Driver State Monitor PASS Fatigue Management System Safe Trac NOV Alert Driver Fatigue Monitor (CoPilot) Voice Commander System Lane Departure EyeCheck Vision/Radar Sensor EDVTCS OSPAT SleepWatch Fatigue Warning System Sleep Control Helmet System Antisleep Insight NapZapper

Company Pernix/FMI Seeing Machines AcuMine Sleep Diagnostics pty Delphi Corporation Advanced Safety Concepts ARRB Transport Research Assistware Technologies Atlas Research ltd Attention Technologies International Mining Technologies Iteris Inc MCJ Mobileye NV Neurocom Ospat Pty Precision Control Design Inc Muirhead/Remote Control tech. Security eEectronic Systems Smart eye SMI Welkin

Parameter of Interest Steering Movement Eye Movements Lane Deviation Eye Movement Eye Blink Detection Head Nodding detection Reaction Time Lane deviation Muscle Tone Analysis Eye Blink Detection Reaction Time Lane Deviation Fitness for Duty Lane Deviation Skin Conductance Fitness for Duty Activity Monitor Reaction Time Head Nodding Eye and Head Movements Eye and Head Movements Head Nodding

Based on table 1. Caterpillar (2008)


15th September 2012 10

Fatigue Detection Technologies

23/10/2012

Medicine, Nursing and Health Sciences

ASTiDTM Advisory System for Tired Drivers

http://www.fmig.org/astid.html .

Taken from www.fmig.org 06/09/2012


Fatigue Detection Technologies 15th September 2012 11

ASTiDTM Advisory System for Tired Drivers


http://www.fmig.org/astid.html The SCIENCE behind key outcome measures
Steering characteristics Forsman et al. 2012 (In press) compared 87 metrics from a high fidelity car simulator:
Only two variables reliably and independently depicted moderate levels of drowsiness Steering Variability (can be converted to lane deviation) Lane Deviation

Knowledge-based characteristics Time of day


e.g. Lenne et al. 1998

Prior sleep
e.g. Horne & Reyner, 1998
Fig: taken from Koh et al. 2007.p 748.

Mabbott & Harley (1998) 50% of night fall asleep events night were preceded by >8h sleep

Time on task
e.g. Phipps-Nelson et al. 2011
Fatigue Detection Technologies 15th September 2012 12

23/10/2012

ASTiDTM Advisory System for Tired Drivers


http://www.fmig.org/astid.html Two integrated systems: Knowledge based Circadian component Quality of prior sleep Time on task factor Steering Sensor Monotonous driving Steering characteristics Generates a Tiredness Score (1-150) Algorithm unknown
Fatigue Detection Technologies 15th September 2012 13

Fig: taken from http://www.fmig.org/astid.html 03/09/2012

Scientific Evaluation

ASTiDTM

Advisory System for Tired Drivers


http://www.fmig.org/astid.html Koh et al. (2007) Construct validity? Criterion (concurrent) validity? Threshold for detection?
Based upon the Sleep Research conclusions of the University of Loughborough [1] a TS of 0.95 is regarded as the threshold for an unacceptably high tiredness level (Koh et al., 2007) [1] = Horne J A and Reyner L A 1995 Sleep related vehicle accidents Br. Med. J. 310 5657

Reliability? Sensitivity/Specificity?

Fatigue Detection Technologies

15th September 2012

14

23/10/2012

Medicine, Nursing and Health Sciences

Seeing Machines Driver State Sensor


http://www.seeingmachines.com

Taken from www.seeingmachines.com 06/09/2012


Fatigue Detection Technologies 15th September 2012 15

Seeing Machines: DSS Driver State Sensor


www.seeingmachines.com The SCIENCE behind key outcome measures
Eye Characteristics -Eye behaviour well known to reflect drowsiness
blink frequency
i.e. Summala et al., 1999

Numerous scientific evaluations Wierwille et al. 1996 PERCLOS and Driving performance
o Correlations between PERCLOS and lane departures/subjective sleepiness SD = lapses, RT, PERCLOS, subjective sleepiness

blink/closure duration
i.e. Anderson et al 2010

Kozak et al. (2005)


o

Delay in lid reopening time


i.e. Caffier et al, 2003

Change in saccadic speed


i.e. Sirevaag & Stern, 2000

PERCLOS
i.e. Dinges et al. 1998

Dinges et al. 1998 PERCLOS and PVT


o o High correlations between PVT lapses and PERCLOS Correlation DECREASED as sleepiness INCREASED.
Fatigue Detection Technologies 15th September 2012 16

23/10/2012

Seeing Machines: DSS Driver State Sensor


www.seeingmachines.com The SCIENCE behind key outcome measures
Eye Characteristics -Eye behaviour well known to reflect drowsiness
blink frequency
i.e. Summala et al., 1999

Numerous scientific evaluations Chau et al. 2012 (p329)


Good inter-individual variation
Correlation with PVT lapses PERCLOS 87% KSS 67% EEG -87% Eye blinks 61%

blink/closure duration
i.e. Anderson et al 2010

Delay in lid reopening time


i.e. Caffier et al, 2003

Change in saccadic speed


i.e. Sirevaag & Stern, 2000

PERCLOS
i.e. Dinges et al. 1998

Sensitivity =78% Specificity = 88%


Fatigue Detection Technologies 15th September 2012 17

Seeing Machines: DSS Driver State Sensor


www.seeingmachines.com The SCIENCE behind key outcome measures
Distraction Characteristics - Sleep loss leads to enhanced distractibility (Anderson & Horne, 2006; Anderson et al., 2010) - Not currently used as a metric for fatigue detection.
Taken from Anderson & Horne, 2006. p 575.

Fatigue Detection Technologies

15th September 2012

18

23/10/2012

Seeing Machines: DSS Driver State Sensor


www.seeingmachines.com Video-based system monitors: Eyelid opening of the driver Face of the driver Metrics clearly shown in literature Taken from http://www.seeingmachines.com/ 03/09/2012 to be sensitive to fatigue.

No known independent validation data to date (03/09/2012)


Fatigue Detection Technologies 15th September 2012 19

Peer reviewed publication following the 2008 fatigue review Three ocular based fatigue detection trialled. NOT IDENTIFIED. Strong links to lane deviation, although for excessive fatigue, the association becomes problematic. Major problems = 1) lack of identifiable outputs 2) Large inter- and intra-individual differences 3) Error rates higher for detecting fatigue compared to EEG
Fatigue Detection Technologies 15th September 2012 20

10

23/10/2012

Scientific Evaluation

Seeing Machines: DSS Driver State Sensor


www.seeingmachines.com Scientific peer-reviewed evaluation Construct validity? Criterion (concurrent) validity? Threshold for detection? Reliability? Sensitivity/Specificity? Note: the review if DSS Construct validity? Criterion (concurrent) validity? Threshold for detection? Reliability? Sensitivity/Specificity?
Fatigue Detection Technologies 15th September 2012 21

Medicine, Nursing and Health Sciences

OPTALERTTM
http://www.optalert.com

Taken from www.optalert.com/ 03/09/2012


Fatigue Detection Technologies 15th September 2012 22

11

23/10/2012

OptalertTM
www.optalert.com The SCIENCE behind key outcomes measures Eyes are a window into the Central Nervous System

Drowsiness leads to a delay in the reopening of an eye closure.


Fatigue Detection Technologies 15th September 2012 23

OptalertTM
Small Light Emitting Diode (500Hz - 940nm). IR light reflected back from the eye. Proprietary software calculates multiple eye and eyelid movement parameters:
1) SD of amplitude-velocity ratio 2) SD of interval between EC/EO

Johns Drowsiness Score (JDS)


0 = VERY ALERT 10 = VERY DROWSY

Barrel transmits and detects light to detect movement of the eye

Fatigue Detection Technologies

15th September 2012

24

12

23/10/2012

OptalertTM
Most scientifically validated system:
Johns, M. et al. Monitoring eye and eyelid movements by infra-red reflectance oculography to measure drowsiness in drivers. Somnologie, 2007; 11:234-242. Johns, M. et al. A new method for assessing the risks of drowsiness while driving. Somnologie, 2008;12: 66-74. Michael, N. et al.effects of caffeine on alertness as measured by infra-red oculography. Psychopharmacology, 2008; 200: 255-260. Johns, M. et al. the effects of blinks and saccadic eye movements on visual reaction times. Attention, Perception and Psychophysics, 2009; 71:783-788. Ftouni,S. et al. Objective and subjective measures of sleepiness, and their associations with on-road driving events in shift workers. Journal of Sleep Research, 2012; Epub. Anderson, C. et al. Assessment of Drowsiness Based on Ocular Parameters Detected by Infra-Red Reflectance Oculography. Under review.
Fatigue Detection Technologies 15th September 2012 25

OptalertTM
Johns, M. et al. 2007; Johns, M. et al. 2008

Alert Driver Reports of Sensitivity/Specificity 83.3% and 60.9% (4 wheels leave carriageway)

Sleepy Driver 75% and 70.6% (2 wheels leave carriageway)


Fatigue Detection Technologies 15th September 2012 26

13

23/10/2012

OptalertTM
Anderson, C. Chang, A-M., Sullivan, J.P., Ronda., J.M. & Czeisler, C.A. 2012. Assessment of Drowsiness Based on Ocular Parameters Detected by Infra-Red Reflectance Oculography. Under Review.
METHODS 29 young, healthy adults: 18-34y (23.3y 4.6 y) 18m; 11f admitted to the CCI at BWH. 30 hour extended wake (constant routine) following an 8hour phase advance protocol Wore OptalertTM throughout Completed a bi-hourly neurobehavioral test battery
Sleep Suite; Centre for Clinical Investigation; Brigham and Womens Hospital, Boston, USA

Fatigue Detection Technologies

15th September 2012

27

Results Analysis 1

Data unpublished.

14

23/10/2012

Results Analysis 2

Data unpublished.

Results Analysis 3

Data unpublished.

15

23/10/2012

OptalertTM

Data unpublished.

Fatigue Detection Technologies

15th September 2012

31

Scientific Evaluation

OptalertTM
Scientific Evaluation Construct validity?
Clear changes in sleep pressure/propensity

Criterion (concurrent) validity?


Associated with current gold standard measures

Threshold for detection?


4.5 5 warning signs for detection

Reliability?
Conference presentation, not yet peer reviewed.

Sensitivity/Specificity?
~ 75% sensitivity and ~60% sensitivity.
Fatigue Detection Technologies 15th September 2012 32

16

23/10/2012

Medicine, Nursing and Health Sciences

SMART CAP
http://www.smartcap.com.au

Taken from www.smartcap.com.au/ 03/09/2012

Fatigue Detection Technologies

15th September 2012 33

Smart Cap
www.smartcap.com.au The SCIENCE behind key outcomes measures Electroencephalography (EEG) clear index of drowsiness. Activities ranging from 4Hz-12Hz clear index of drowsiness Clear transient signs of sleep The Gold Standard measure drowsiness

Cajochen et al. 1999. Ocular & EEG correlates over 40h awake.

Taken from Cajochen et al. (1999). P.R644


Fatigue Detection Technologies 15th September 2012 34

17

23/10/2012

Smart Cap
www.smartcap.com.au The SCIENCE behind key outcomes measures Electroencephalography (EEG) clear index of drowsiness. Activities ranging from 1Hz-12Hz clear index of drowsiness Clear transient signs of sleep The Gold Standard measure drowsiness
Taken from Cajochen et al. (1999). P.R645

Cajochen et al. 1999. Ocular & EEG correlates over 40h awake.

The crosscorrelation analyses between EEG power density in selected frequency bands and psychomotor vigilance performance demonstrate that frontal EEG power density in the slowwave and theta bands are most highly correlated with changes in performance.
15th September 2012 35

Fatigue Detection Technologies

Scientific Evaluation

SmartCap
Baseball cap (headband/visor) with concealed sensors which monitor an operators brain waves (EEG) to calculate a measure of drowsiness Information is wirelessly communicated to a display in-cab, or to any Bluetooth enabled device. Developed within CRCMining, a Cooperative Research Centre established by the Australian government, supported by four universities and 13 industry partners including equipment manufacturers and mining companies. Award winning
Australian mining prospect award (2009) Finalist for Pace/Zenith Award (2010) CRC Association Award for Excellence in innovation (2011) Australasia Young Public Safety Award (2012)

Scientific peer-reviewed validation needed (likely imminent)


Fatigue Detection Technologies 15th September 2012 36

18

23/10/2012

Fatigue Detection Technologies: FUTURE


Future technologies will likely focus on more than one attribute European Union AWAKE project
System for Effective Assessment of Driver Vigilance and Warning According to traffic Risk Estimation
Eyelid movements Steering grip Driver behaviour (lane tracking, steering, braking)

SAVE Project
System for Effective Assessment of the Driver State and Vehicle Control in Emergency Situations
Physiological measures (eyelid movements; head position; grip force) Vehicle measures (speed; steering, distance to objects; lateral position) Environmental measures (time of day; weather)

Currently, detects ~90% fatigue (Mabbott et al., 1999)


Fatigue Detection Technologies 15th September 2012 37

Fatigue Detection Technologies: FUTURE


Future technologies will likely focus on more than one attribute DROWSI Project (Drowsiness Intervention Project)
DROWSI detection system

blink behavior, steering wheel control and lateral variability Future technologies All measures are some time off being commercially available.

Fatigue Detection Technologies

15th September 2012

38

19

23/10/2012

Current Fatigue Technologies


Take Home Message Fatigue detection technologies are an integral part of fatigue management Two levels of validation required for technology: Scientific validation Operational/ecological validation Four current technologies based on strong scientific outcomes Scientific validation of fatigue detection inc. threshold sensitivity is lacking Beyond technology: who sees the results? Are employees penalised or rewarded
for sleep-related behaviours; what warning signals are most effective; when are warning signals effective; what level of false positive is acceptable; impact on driver workload.

Moving forward: combining scientific evaluation with operational/industry input is critical for reducing fatigue through technology based intervention.
Fatigue Detection Technologies 15th September 2012 39

Medicine, Nursing and Health Sciences

Fatigue Detection Technologies


Clare Anderson, Ph.D. 1,2,3
1Sleep 2

& Chronobiology Research group, Monash University Brigham and womens Hospital, Boston, USA. 3 Harvard Medical School, Boston, USA

QUESTIONS?

20

Anda mungkin juga menyukai