functional assessment of head–eye coordination during ... · original article functional...

ORIGINAL ARTICLE

Functional Assessment of Head–EyeCoordination During Vehicle Operation

HAMISH G. MACDOUGALL, PhD, and STEVEN T. MOORE, PhD

Department of Neurology, Mount Sinai School of Medicine, New York, New York

ABSTRACT: Purpose. Visual impairment, resulting from ocular abnormalities or brain lesions, can significantly affect drivingperformance. The impact of vestibulopathy on head–eye coordination is also a concern in vehicle operation safety, yet todate there has been little functional research in this area. An understanding of decrements in driving ability resulting fromvisual and vestibular pathology, plus the differences in visual strategies used by novice and experienced drivers, wouldbenefit from an objective analysis of head–eye coordination during vehicle operation. Methods. We have developed alaptop-based system for measuring eye, head, and vehicle movement in real time. Digital video cameras mounted onlightweight swimming goggles are used to provide images of the eye and scene, allowing assessment of gaze. In addition, theuse of inertial measurement units to simultaneously transduce head and vehicle movement allows us to evaluate thevestibular contribution to stable vision. Results. Data was obtained from a flight simulator and while driving a car. Duringbanking turns in the flight simulator, there was a sustained roll tilt of the head and eyes toward the scene-derived visualvertical with a combined gain of approximately 25%. One of the most complex visual tasks when driving was exiting amultistory car park, which involved the scanning of hundreds of parked vehicles with an average fixation time ofapproximately 100 ms. The vertical vestibulo-ocular reflex was also found to make a significant contribution to themaintenance of dynamic visual acuity even while driving on paved surfaces. Conclusion. These results demonstrate theviability of functional assessment of head–eye coordination during vehicle operation, and potential applications of thistechnology to driver assessment are discussed. Analysis of both active and reflex contributions to gaze may provide a clearerunderstanding of the impact of visual and vestibular impairment on driving ability. (Optom Vis Sci 2005;82:706–715)

Key Words: head tracking, eye movements, vestibular, driving, flight simulator

Motor vehicle accidents account for 47% of accidentaldeaths1 and 94% of all transportation-related fatalities inthe United States,2 and are the leading cause of death in the

1- to 33-year age group.1 The distribution of fatalities by age is bi-modal, with both young (age range, 15–24 years) and older (75�years) groups being most affected, and consequently much researchhas focused on the effects of the inexperience of novice drivers3-5; andon the sensorimotor, cognitive, and visual deficits associated with ag-ing (see Wood6 for an extensive review). Driving is a vision-intensivetask, and recent studies have demonstrated that age-related decre-ments in vision increase the probability of an accident, as determinedfrom a statistical analysis of visual ability and driving records,7-9 andfunctional studies of road sign and hazard recognition on a closed-roadcircuit.10 Visual pathology is unlikely to be a major contributing factorto the increased fatality rate for the younger (age range, 15–24 years)group of drivers, but recent open-road studies have identified a morelimited pattern of gaze fixations in novice drivers as compared withthose with more experience.3,4 Thus, both visual function and thescanning techniques learned by drivers are important aspects of con-trolling an automobile.

Gaze analysis is a useful tool in understanding the visual behav-iors underlying driving such as fixation of the tangent point whennegotiating a curve,11,12 the near (9-m) and far (16-m) points ofregard for maintaining position in lane and assessment of roadcurvature,12 the wider distribution of horizontal gaze movementsin drivers of large vehicles at intersections,13 and differences in thescanning patterns of novice and experienced drivers.3,4 All thesestudies have used some combination of eye-in-head tracking andhead-referenced scene images to analyze where the driver was look-ing; including manual analysis of videotaped images of the eye11,12

and a commercial eye tracker (Eyemark VII, NAC Inc., Tokyo,Japan) measuring reflection of an infrared (IR) light source off thesclera (IR oculography [IROG]).3,4 However, these approacheshave limitations. Manual processing is clearly a time-consumingtask,a and the IROG technique is inaccurate, prone to thermaldrift, and essentially restricted to horizontal eye movements,14 al-lowing only a general description of where a driver is looking.

Clinical assessment of eye position has commonly used electro-

aOne minute of videotape comprises 1500 (PAL) to 1800 frames (NTSC).

1040-5488/05/8208-0706/0 VOL. 82, NO. 8, PP. 706–715OPTOMETRY AND VISION SCIENCECopyright © 2005 American Academy of Optometry

Optometry and Vision Science, Vol. 82, No. 8, August 2005

oculography (EOG), in which surface electrodes measure variations inthe electrical potential of the eye, and this technique has been used inflight simulator studies of pilot performance.15 EOG, however, hassimilar drawbacks to the IROG technique.14 Advances in image pro-cessing hardware has led to the development of automated noninva-sive techniques for measurement of eye position using images fromvideo cameras (video oculography [VOG]).16-27 Current commercialVOG systems use analog video cameras, requiring desktop computerswith video acquisition cards to process eye movements, limiting theiruse in a real-world setting (e.g., Vision 2000, El Mar, Toronto, Can-ada; iView X, SMI, Teltow, Germany; ISCAN, Cambridge, MA).Moreover, the head-mounted cameras add significant mass to thehead (300-g Vision 2000; 450-g iView X), which may impact opera-tor behavior. Commercial VOG systems incorporating a scene camerato study gaze patterns have recently been used in flight simulatorstudies28 and for monitoring the gaze of pedestrians crossing an inter-section29 (recording images to VCRs in the subject’s backpack) andwalking through a complex indoor environment30 (with a 30-m um-bilical cable).

An often neglected aspect of viewing behavior while driving ismovement of the head. Two on-road studies have investigatedone-dimensional head movement: head yaw (estimated from theposition of a radio antenna in the scene image) in racing car driversshowed that subjects “steered” with their heads by pointing thenaso-occipital axis into the turn,31 much like humans walkingaround a corner,32 and the magnitude of roll (lateral) tilt of thehead into a turn (measured with an accelerometer) was related toroad curvature.33

Head movement is particularly relevant to the study of vestibu-lar-related driving difficulties. The peripheral vestibular labyrinthssense angular and linear head movement, and this information iscentrally processed to provide postural stability, compensatory eyemovements (the vestibulo-ocular reflex), and awareness of bodyposition in space (spatial orientation). Damage to the vestibularsystem may occur after head injury,34 viral infection of the vestib-ular nerve,35 minor strokes involving the anterior inferior cerebel-lar artery,36 use of ototoxic antibiotics (such as gentamicin),37 andsurgical procedures. Age-related decrements in vestibular functionare also well-documented,38,39 likely related to degeneration atboth the peripheral 40-42 and central43 level. Vestibulopathy is typ-ically manifested as vertigo (sensation of spinning) and disturbedvision (inappropriate nystagmus, oscillopsia), often provoked bysudden head movement. For example, benign paroxysmal posi-tional vertigo (BPPV), a common cause of dizziness,44 is thoughtto occur when small crystals of calcium carbonate (otoconia) fromthe utricular macula (the “linear accelerometers” of the vestibularlabyrinth) migrate into the semicircular canal system. Changes inhead position with regard to gravity, particularly in pitch or roll,trigger movement of the otoconia within the canal, generatingvertigo and ocular nystagmus. In the acute phase of recovery aftervestibular nerve sections or acoustic neuroma resections, patientsoften have brief episodes of vertigo during head rotations withassociated blurred vision.45 In contrast, Meniere’s disease is char-acterized by sudden and unexpected attacks of vertigo usually un-related to head movement.46

Vestibular impairment engenders a substantial reliance on visualinformation to maintain balance and gaze, and a suppression ofhead movement to prevent vertigo and inappropriate ocular nys-

tagmus, which may affect the ability of the individual to drive.Moreover, the vestibulo-ocular reflex (VOR), which acts to stabi-lize gaze by generating compensatory eye movements in the oppo-site direction to head movement, is critical in maintaining dynamicvisual acuity while driving, because even the paved roads of urbancenters generate high-frequency perturbations of the head. Clini-cians in a number of countries have expressed significant concernwith regard to vestibular disease and driving, 47-50 although to date,there has been little functional research in this area. A recent studyby Cohen et al45 assessed the impact of vestibular dysfunction ondriving performance from subjective reports using a modified formof the Driving Habits Questionnaire developed for patients withvisual impairment.8 The subject pool represented a broad spec-trum of vestibular disorders, including BPPV, postoperative acous-tic neuroma resections and vestibular nerve sections, Meniere’sdisease, and chronic idiopathic vestibulopathy. These subjects re-ported considerable difficulty driving in reduced visibility (such asat night or during rain) and in visually complex environments(high-traffic roads, large intersections). Complex maneuvers in-volving spatial navigation such as changing lanes, staying in lane,and parking were problematic, likely as a result of impairment ofpath integration (i.e., summing of self and vehicle movement overtime) that is degraded in vestibular patients.51 Vertigo and ocularnystagmus were also triggered by tasks requiring rapid head mo-tion such as checking for traffic before changing lanes or enteringan intersection or freeway on-ramp.45 The impact on stable visioncan be devastating, because inappropriate nystagmus (i.e., notcompensatory for head rotation) causes “spinning” of the visualscene.

Head–eye coordination is also critical in assessing the impact ofvisual field deficits as a result of ocular abnormalities52,53 (maculardegeneration, glaucoma, or retinitis pigmentosa) or brain lesions(such as hemianopsia, a loss of vision in half of the visual field oftenrelated to stroke,54-56 or head trauma). Visual field loss does notaffect visual identification but alters search strategies,57 with longerscan paths, more frequent and prolonged fixations, and conse-quently increased error.58 Transportation authorities in the UnitedStates and Europe typically require a minimum 120° horizontalfield of view in the better functioning eye to obtain a drivinglicense. However, recent studies have suggested that many patientswho do not meet this criterion are able to drive safely with appro-priate adaptation of head–eye coordination. Of 35 subjects withperipheral visual field deficits resulting from ocular disease (with ahorizontal field extent limited to 84° on average), 15 (43%) passedan on-road driving assessment.52 Furthermore, patients whopassed the test were found to have performed a greater number ofhead movements and began scanning earlier when approaching anintersection than those who failed. A similar study was recentlyperformed with hemianopic patients whose visual field is typicallylimited to 90°.59 Training patients to perform frequent, large hor-izontal saccades into the blind hemifield, while minimizing headmovement, generated a subjective enhancement of vision and mea-surable improvement in driving ability.59

Analysis of head–eye coordination and vestibular contributionsto visual behavior while driving is important in establishing stan-dard operator performance, and for assessment of drivers withvestibular and visual field impairment. In this article, we describe anovel laptop-based system for simultaneous acquisition of three-

Head and Eye Coordination in Driving—MacDougall and Moore 707


dimensional (3D) head, eye, and vehicle movement in real time.The system is comprised entirely of commercially available hard-ware and uses established VOG algorithms to accurately determinehorizontal and vertical movements of the pupil, as well as rotationof the eye around the line of sight (ocular torsion).20,21 The port-ability of the system allows for the functional study of head–eyecoordination and the vestibulo-ocular reflex in situations that werenot previously possible, and as examples, we provide data obtainedfrom a pilot operating an Airbus A340-600 motion simulator andfrom a subject driving a car in Manhattan.

METHODS

The experiments described subsequently were approved by theInstitutional Review Board of the Mount Sinai School of Medicine

and conformed to the Declaration of Helsinki. Informed consentwas obtained from all subjects.

Hardware

The head–eye tracker was developed as part of a NASA-fundedstudy of spatial disorientation in shuttle pilots60 and was imple-mented using commercially available components. Two “firewire”(IEEE 1394) digital cameras (Firefly; Point Gray Research, BC,Canada) were attached to lightweight swimming goggles (Aquas-phere Seal, Genova, Italy) (Fig. 1A,B). The total weight of theheadset was 146 g, significantly lighter than commercially availableVOG goggles. A limitation of the current system is that spectaclescannot be worn under the goggles (although contact lenses can beused at the expense of torsional eye position measurement accu-

FIGURE 1.A laptop-based video oculography (VOG) system for assessment of head–eye coordination during driving. (A) The subject wore goggle-mounted digitalvideo cameras that imaged the left eye and the scene. An inertial measurement unit (IMU) was attached to the head band to measure head movementin space. The total weight of the headset was 146 g, which is considerably lighter than commercial VOG systems. (B) A second IMU was mounted tothe vehicle to measure motion of the car in space. The video cameras and IMUs were connected to a laptop computer that provided eye, head, andcar movement data and gaze-in-scene in real time, as well as generating digital video files for later display. (C) Screen dump from the laptop displayduring acquisition of data while driving in New York City. Color versions of Figures 1 through 6 are available at www.optvissci.com.

708 Head and Eye Coordination in Driving—MacDougall and Moore


racy), but the use of a head-band arrangement to mount the cam-eras resolves this issue. The left eye was illuminated by an IRlight-emitting diode (HSDL-4220; Hewlett-Packard, Houston,TX), and the image of the eye was directed onto the left camerathrough a dichroic mirror (Wideband Hot Mirror, OCLI, CA)that was sensitive to light in the IR band but allowed visible light topass through, providing a clear field of view. A second camera wasdirected along the naso-occipital axis, providing images of thescene from the subject’s point of view. Note that this camera couldalso be used to obtain binocular images of the eyes. The cameracables (Apple Thin Firewire Cable; Apple Computer Inc., CA)were connected to a PCMIA firewire card (IEEE-1394 CardBusPC Card CBFW3U; Ratoc Systems International, San Jose, CA)inserted into the PC slot of the laptop computer (SONY VaioPCG-GRV670P) and powered from the laptop battery. The use ofdigital rather than analog video cameras eliminated the need for avideo acquisition card, because digital images of the eye were pro-vided directly from the cameras.

Eye movements were calibrated by having the subject view tar-gets generated by a goggle-mounted laser (class 3A �5 mw 635 nmvisible laser diode module) with a diffraction grating (Laser X-HairGenerating Optic, Lasermate Group Inc., Pomona, CA), whichprojected a “cross” pattern of lines that subtended a visual angle of�11° horizontally and vertically and were fixed with respect to thehead. Head and vehicle movement were determined with inertialmeasurement units (IMUs), connected to the laptop USB port,which used triaxial accelerometers and angular rate sensors to mea-sure 3D linear acceleration and angular velocity in space. In addi-tion, yaw, pitch, and roll orientation of the head and vehicle inspace was calculated from the IMU linear acceleration and angularvelocity data using three integral flux gate compasses and temper-ature compensation to suppress drift (MT9; Xsens, Enschede, TheNetherlands). Addition of a USB analog/digital data acquisitioninterface (PMD-1208LS USB-based DAQ module; MeasurementComputing Corp., MA) allowed auxiliary analog or digital data tobe collected simultaneously. As of January 2005, the total cost ofthe hardware was less than U.S. $6000.

Software

The system software was written in Labview G (National Instru-ments, Austin, TX). Images of the left eye and scene were acquired ata rate of 30 Hz,b and the center of the pupil was determined using a“center-of-mass” algorithm.20,21 Horizontal and vertical eye positionwere calculated in Fick coordinates using a spherical model of theeye,21 the radius of which was calculated using a calibration procedurein which the subject fixated targets at known gaze angles generated bythe head-referenced laser. The laser display was also visible in the sceneimage to calibrate gaze, and total setup time was �5 min. Torsionaleye position was calculated using polar crosscorrelation,16,18-21 inwhich pixels within the iris are sampled along elliptical annuli centeredon the pupil and crosscorrelation of these signals provides the amountof relative rotation about the line of sight between two images.20–22

These algorithms have demonstrated an accuracy and resolution of theorder of 0.1°.21

The three dimensions of eye position (horizontal, vertical, andtorsional), head and vehicle movement, and analog or digital datalogged with the data acquisition card were presented in real time onthe laptop screen (Fig. 1C). In addition to providing eye positiondata, the point of regard was superimposed on the correspondingscene image to allow analysis of patterns of gaze in real time (Fig.1C). Although commercial desktop-based VOG systems withscene cameras have been available in recent years, our system is thefirst to integrate 3D eye-in-head position, 6 degrees-of-freedomhead and vehicle movement, plus gaze-in-scene in real time withina laptop-based unit suitable for use in vehicles and simulators.

RESULTSFlight Motion Simulator

The system was used to acquire eye, head, and cabin movementdata aboard a full flight motion simulator (A340-600) at the Air-bus training facility in Toulouse, France. Figure 2 shows roll headand eye movements during sustained (30 s) 45° banking turnsduring a fixed-base simulator run that modeled the heading align-ment circle (HAC) maneuver during the final approach of thespace shuttle.60 In response to the tilt of the visual horizon, therewas a maintained tilt of the head of up to 5° (Fig. 2: lower panel,solid line), termed the optokinetic cervical reflex,61,62 although inthis instance, it is likely a combined optokinetic and optostaticresponse. In addition, there was a sustained torsional shift in eyeposition of 6° (Fig. 2: lower panel, dashed line), which precededthe head tilt. This ocular torsion was of similar magnitude to thatproduced by the gravity-sensing otoliths during a 45° head tilt withregard to gravity (ocular counterrolling [OCR]).63 It is importantto note, however, that the OCR reflex would rotate the eye in theopposite direction to head tilt, whereas in this instance, both thehead and eye rotated toward the scene-derived “visual vertical”(i.e., perpendicular to the horizon). Ocular torsion in response to arotating visual line has recently been described,64 and this is, to ourknowledge, the first account of sustained ocular torsion in responseto a statically tilted scene. The combined head and eye roll tilt actedto orient the eye to the scene-derived visual vertical with a gain ofapproximately 25%.

Driving a Car

Data were obtained from an experienced subject (licensed 23years) driving a midsized four-door sedan (2000 Chevrolet Cava-lier LS) in daylight in a dense, urban environment (Manhattan).Figure 3 shows a sequence of images as the driver negotiated a seriesof 180° turns in a multilevel car park. The analysis software was setto output gaze fixations �200 ms either side of the current sceneimage to allow complex scanning patterns to be represented in astatic image. The subject scanned the reverse and brake lights onthe rear of each parked vehicle ahead in a linear manner with afixation time of approximately 100 ms per vehicle (i.e., on averagefour cars were scanned in each panel of Fig. 3, which represents aperiod of 400 ms). When a vehicle partially reversed into his path,the driver maintained fixation on the lit reverse indicator for alonger period (Fig. 3: panels 5 and 6) while continuing to scan thecars ahead. A similar scanning strategy continued when turningright from the car park onto the street (Fig. 4A). Other scanning

bPreliminary trials of a new generation of firewire cameras (Scorpion, Point GreyResearch, BC Canada) demonstrated a sample rate of 150 Hz.



behaviors included checking the position of pedestrians while pre-paring to make a righthand turn (Fig. 4B), estimating the gapbetween two parked trucks before passing (Fig. 4C), reading anoverhead road sign (Fig. 4D), fixating alternately on the tangentpoint of a curving road11,31 and the car ahead (Fig. 4E), scanningthe rearview mirror and the road ahead (Fig. 4F), and checking thelefthand side mirror and a van in the driver’s blind spot beforechanging lanes to the left (Fig. 4G, H).

A novel aspect of our system is that it seamlessly provides syn-chronous head and vehicle movement data to augment the eye-in-head position data and gaze-in-scene. As the driver negotiated atwisting off-ramp from the George Washington Bridge (Fig. 5),

both the fixation of the tangent of the curve and the roll tilt of thehead into the turn are clearly shown, as described in two previousseparate studies.12,33 The tilt of the head averaged approximately10° into the turn (Fig. 5: lower panel, solid line), which was ap-proximately half the angle of the gravito-inertial accelerationc

(GIA) tilt (Fig. 5: lower panel, dashed line). The magnitude ofhead and GIA tilt were similar to that observed in humans walkingaround a 0.5-m radius turn.32

During unpredictable passive head movement, the vestibulo-

cThe vector sum of gravity and the centripetal acceleration of the car as itrounded the bend.

FIGURE 2.Sequence of six head-referenced scene images from a pilot flying an Airbus A340-600 simulator (in this instance, fixed-base). The subject performedsustained (30 s) 45° banking turns to the left and right to simulate the final approach of the space shuttle. The superimposed cross shows the point ofregard within the scene, and the orientation of the cross represents the torsional component of eye position in head coordinates (with a gain of 10). Thehead image below each scene shows the corresponding tilt of the head in space (as viewed from behind the subject). The graph at the bottom of thefigure shows head-in-space and eye-in-head roll tilt data (in degrees), and the numbered cursors indicate the relative temporal location of the six frames.Both the head and eye rolled about the naso-occipital axis toward the scene-derived visual vertical with a combined gain of 25%.



ocular reflex (VOR) stabilizes gaze through compensatory eyemovements. Figure 6 shows 15 s of head and eye pitch (rotationabout axes parallel to the interaural axis) while driving along West96th Street on Manhattan’s upper west side. Although this was apaved road in reasonably good condition, there was a continuoushigh-frequency pitching of the head (with peak-to-peak amplitudeof approximately 6°) as a result of the vertical movement of theautomobile over the uneven road surface (Fig. 6A: solid line). Thepitch of the head, sensed by the semicircular canals, was used togenerate vertical eye movements by the VOR that were of similaramplitude but in the opposite direction to head movement tomaintain gaze (Fig. 6A: dashed line). A scatterplot of eye versushead position demonstrates this compensatory reflex with a linearregression showing a slope of close to unity (0.9) (Fig. 6B). Thus,the VOR is constantly generating compensatory eye movement toovercome passive perturbations of the head, which augments activegaze fixations to maintain stable vision when driving.

DISCUSSION

Driving a car is a visually complex undertaking, requiring inte-gration of both active and reflexive head and eye movements toprovide a stable view of the surroundings. Human head—eye co-ordination did not evolve to suit the high-inertial environment of

powered vehicles, and these patterns of behavior must be learned,as evidenced by the limited scanning strategies of novice drivers.3,4

Moreover, visual impairment, which may not greatly affect activelocomotion, can significantly diminish the acquired skill of auto-mobile control.7,8,10,45 Obtaining functional measures of head–eye coordination during vehicle operation is an important aspect inunderstanding the basic visual strategies underlying driving, as wellas assessment of the impact of visual and vestibular impairment ondriving performance. Our laptop-based system allows the real-timeanalysis of 3D head–eye coordination that incorporates both ac-

FIGURE 3.Sequence of six scene images from a subject driving in a multistoryparking garage. The sequence of fixations 200 ms either side of eachimage are represented as dots. The subject fixated for approximately 100ms on the reverse lights on the rear of each parked car ahead in a linearmanner, checking for reversing vehicles. In frame 5, an SUV partiallyreversed out into the cars path, and the driver continued to monitor itsreverse indicator while continuing to check the cars further ahead (frame6).

FIGURE 4.Scene images from a subject driving around Manhattan. The sequence offixations 200 ms either side of each image are represented as dots. (A) Thesubject scanned the brake lights and wheels of parked vehicles to checkfor potential hazards when exiting the car park onto a city street. (B) Beforemaking a righthand turn at traffic lights, the subject scanned the move-ment of three pedestrians with the potential to intrude on the path of thevehicle. (C) When approaching two parked trucks, the driver scanned thegap between them to determine there was adequate room to pass through.(D) The subject fixated on a directional road sign before exiting a freeway.(E) On the curving approach to the George Washington Bridge, the subjectfixated both on the tangent of the curve and a vehicle ahead. (F) On thebridge, the driver scanned the rear vision mirror and the road ahead. (G)The driver checked the leftside mirror before a lane change to the left, and(H) then turned his head to check the blind spot and fixated on a van inthe left lane.



tive head and eye fixation strategies, plus reflex contributions fromthe vestibulo-collic and vestibulo-ocular reflexes. The accuracy ofthe system is such that complex visual behaviors can be identifiedand quantified with a high temporal and spatial resolution.

To date, there has been little emphasis on the vestibular aspectsof driving, at least in the functional sense presented here, but this isclearly an important issue. The VOR is continuously operationalwhile driving, underpinning the active visual scanning strategiesdescribed here. Although patients with vestibulopathy may over-come decrements in VOR performance to some extent with pur-suit and saccadic eye movements, this strategy breaks down in poorvisual conditions or complex environments and during rapid headmotion.45 As an example, the seemingly straightforward task ofexiting a car park involves the methodical scanning of parked carsat rates of up to 10 per second (Fig. 3), which must be accom-plished during rapid turns of the vehicle and the accompanyingvestibular-generated head and eye movements to provide a seam-less view of an environment with many potential hazards. A largenumber of drivers with vestibular impairment report significantdifficulty in negotiating multistory car parks.45 Inappropriate oc-

ular nystagmus, and therefore paradoxic motion of the visual sur-round, can be induced by activation of the semicircular canalsduring turns of the head and vehicle. In addition, the visual struc-ture of car parks, which often have evenly spaced vertical columnsor blinds to allow natural light, can present a challenge. Unfortu-nately, for drivers with vestibulopathy, the horizontal motion ofthese vertical structures while turning are akin to a rotating opto-kinetic drum45 and can themselves induce episodes of vertigo andoptokinetic nystagmus.d

The integration of gaze analysis with head and vehicle move-ment allows the evaluation of vestibular reflexes that maintainstable vision while driving. This may prove a useful adjunct topurely subjective forms of assessment of vestibulopathy on drivingability45 as well as addressing the larger question of what level ofvestibular impairment is sustainable for safe driving.47–50 For thecar park example presented here, episodes of vertigo with accom-

dThe optokinetic reflex generates ocular nystagmus in response to movement ofthe visual surround and shares many of the same neural pathways as the vestibulo-ocular reflex.

FIGURE 5.A sequence of three scene images while negotiating a series of curves exiting the George Washington Bridge into Manhattan. The sequence of fixations200 ms either side of each image are represented as dots. The head image below each scene shows the corresponding tilt of the head in vehiclecoordinates (as viewed from behind the subject). The car image below the head shows the vector sum of gravity and the centripetal acceleration of thevehicle (the gravito-inertial acceleration [GIA]). The graph at the bottom of the figure shows head and GIA roll tilt data in vehicle coordinates, and thenumbered cursors indicate the relative temporal location of the three frames. When cornering (frames 1 and 3), the driver fixated on the tangent of thecurve. There was a large tilt of the GIA of approximately 20°, and a corresponding 10° roll tilt of the head, into the curve.



panying nystagmus can be identified and linked to specific head orvehicle movements or features of the visual environment. A similarapproach could be used to establish driving maneuvers involvinghead motion that are of particular concern such as changing lanes(Fig. 4H), cornering (Fig. 5), and driving on rough road surfaces(Fig. 6), and potentially lead to modifications in driving behaviorto limit disorientation. This approach is used in vestibular rehabil-itation such as advising patients with BPPV to avoid placing ob-jects on high shelves in the home to minimize head pitch move-ments that trigger vertigo. It is interesting to note that of the fourvestibular patient populations in the Cohen study,45 those withBPPV reported the lowest incidence of driving difficulty. Becausetheir vertigo is linked to specific active head movements in the rollor pitch plane, it is likely that these patients have adapted theirhead–eye coordination strategy to circumvent an on-road attack.A functional assessment of head and eye movement during drivingwould be a useful means to investigate this hypothesis.

The dangers of driving with poor vision are well document-ed,7,8,10 and although our system cannot determine how well adriver sees the world, it can be used to assess the effects of oculardegeneration on the visual performance of drivers in a functionalsetting. Analysis of head–eye coordination could be useful in as-sessment of drivers with visual neglect or visual field deficits, whichrequire a change in scanning strategies to improve vision.58,59,65,66

Recent studies have suggested that adaptation of head–eye coor-dination by patients with visual field deficits can improve vehicleoperation,52,59 even to the extent of passing an on-road driving

assessment,52 despite the fact that the horizontal visual extent inthese individuals officially precludes them from obtaining a li-cense. Thus, training patients to adapt head–eye coordination toextend their functional horizontal field, possibly in conjunctionwith the use of auxiliary optical aids (such as prisms65,66), mayimprove the quality of life of many individuals currently deemedunfit to drive.

A detailed analysis of head–eye coordination and visual searchstrategies may also be of use in driver training. The more sophisti-cated scanning patterns of experienced drivers3,4 are presumablylearned over many years on the road, and this information could beused in the instruction of novice drivers. This objective approachto teaching vehicle control based on the visual strategies of experi-enced operators has been evaluated by the U.S. Air Force67 and iscurrently being investigated by commercial aviation and NASAusing the apparatus described here.60

ACKNOWLEDGMENTS

Supported by NASA grants NCC 9–128 and NNJ04HF51G (Steven Moore)and a National Space Biomedical Research Institute (NSBRI) postdoctoralfellowship (Hamish MacDougall).

Received November 30, 2004; accepted February 4, 2005.

REFERENCES

1. National Safety Council. Injury Facts: 2003. Itasca, IL: NationalSafety Council; 2003.

2. US Department of Transportation Bureau of Transportation Statis-tics. National Transportation Statistics 2002. Washington, DC: USGovernment Printing Office; 2002. Available at: http://www.bts.gov/publications/national_transportation_statistics/2002/.Accessed April 25, 2005.

3. Crundall DE, Underwood G. Effects of experience and processingdemands on visual information acquisition in drivers. Ergonomics1998;41:448–58.

4. Underwood G, Chapman P, Brocklehurst N, Underwood J,Crundall D. Visual attention while driving: sequences of eye fixationsmade by experienced and novice drivers. Ergonomics 2003;46:629–46.

5. Wood JM, Mallon K. Comparison of driving performance of youngand old drivers (with and without visual impairment) measured dur-ing in-traffic conditions. Optom Vis Sci 2001;78:343–9.

6. Wood JM. Aging, driving and vision. Clin Exp Optom 2002;85:214–20.

7. Owsley C, Ball K, McGwin G, Sloane ME, Roenker DL, White MF,Overley ET. Visual processing impairment and risk of motor vehiclecrash among older adults. JAMA 1998;279:1083–8.

8. Owsley C, Stalvey B, Wells J, Sloane ME. Older drivers and cataract:driving habits and crash risk. J Gerontol A Biol Sci Med Sci 1999;54:M203–11.

9. Owsley C, Stalvey BT, Wells J, Sloane ME, McGwin G Jr. Visual riskfactors for crash involvement in older drivers with cataract. ArchOphthalmol 2001;119:881–7.

10. Wood JM. Age and visual impairment decrease driving performanceas measured on a closed-road circuit. Hum Factors 2002;44:482–94.

11. Land MF, Lee DN. Where we look when we steer. Nature 1994;369:742–4.

12. Land M, Horwood J. Which parts of the road guide steering? Nature1995;377:339–40.

FIGURE 6.(A) Head and eye pitch movements while driving along a city street. Therewas a continuous high-frequency perturbation of the head (solid line). Thevestibulo-ocular reflex (VOR), mediated by the semicircular canals, gen-erated compensatory vertical eye movements to maintain gaze (dashedline). (B) Scatterplot of eye versus head pitch. A linear regression analysisdemonstrates the compensatory nature of the VOR with a slope close tounity (0.9). Drivers with vestibulopathy would likely have a limited VORresponse, thus the compensatory eye movements would not be sufficient,resulting in oscillopsia and reduced dynamic visual acuity.



13. Kito T, Haraguchi M, Funatsu T, Sato M, Kondo M. Measurementsof gaze movements while driving. Percept Mot Skills 1989;68:19–25.

14. Young LR, Sheena D. Eye-movement measurement techniques. AmPsychol 1975;30:315–30.

15. Samel A, Diedrich A, Drescher J, Lorenz B, Plath G, Vejvoda M,Wenzel J. [Long-term monitoring of psychophysiologic values inflight physiology.] Internist (Berl) 1997;38:755-69.

16. Hatamian M, Anderson DJ. Design considerations for a real-timeocular counterroll instrument. IEEE Trans Biomed Eng 1983;30:278–88.

17. Parker JA, Kenyon RV, Young LR. Measurement of torsion frommultitemporal images of the eye using digital signal processing tech-niques. IEEE Trans Biomed Eng 1985;32:28–36.

18. Clarke AH, Teiwes W, Scherer H Videooculography: an alternativemethod for measurement of three-dimensional eye movements. In:Schmid R, Zambarbieri D, eds. Oculomotor Control and CognitiveProcesses: Normal and Pathological Aspects. Amsterdam: North-Holland; 1991:431-43.

19. Clarke AH, Ditterich J, Druen K, Schonfeld U, Steineke C. Usinghigh frame rate CMOS sensors for three-dimensional eye tracking.Behav Res Methods Instrum Comput 2002;34:549–60.

20. Moore ST, Curthoys IS, McCoy SG. VTM–an image-processingsystem for measuring ocular torsion. Comput Methods ProgramsBiomed 1991;35:219–30.

21. Moore ST, Haslwanter T, Curthoys IS, Smith ST. A geometric basisfor measurement of three-dimensional eye position using image pro-cessing. Vision Res 1996;36:445–59.

22. Haslwanter T, Moore ST. A theoretical analysis of three-dimensionaleye position measurement using polar cross-correlation. IEEE TransBiomed Eng 1995;42:1053–61.

23. Bos JE, de Graaf B. Ocular torsion quantification with video images.IEEE Trans Biomed Eng 1994;41:351–7.

24. Groen E, Bos JE, Nacken PF, de Graaf B. Determination of oculartorsion by means of automatic pattern recognition. IEEE TransBiomed Eng 1996;43:471–9.

25. Imai T, Takeda N, Morita M, Koizuka I, Kubo T, Miura K, Naka-mae K, Fujioka H. Rotation vector analysis of eye movement in threedimensions with an infrared CCD camera. Acta Otolaryngol 1999;119:24–8.

26. Zhu D, Moore ST, Raphan T. Robust pupil center detection using acurvature algorithm. Comput Methods Programs Biomed 1999;59:145–57.

27. Zhu D, Moore ST, Raphan T. Robust and real-time torsional eyeposition calculation using a template-matching technique. ComputMethods Programs Biomed 2004;74:201–9.

28. Cheung B, Hofer K. Eye tracking, point of gaze, and performancedegradation during disorientation. Aviat Space Environ Med 2003;74:11–20.

29. Geruschat DR, Hassan SE, Turano KA. Gaze behavior while crossingcomplex intersections. Optom Vis Sci 2003;80:515–28.

30. Vivekananda-Schmidt P, Anderson RS, Reinhardt-Rutland AH,Shields TJ. Simulated impairment of contrast sensitivity: perfor-mance and gaze behavior during locomotion through a built environ-ment. Optom Vis Sci 2004;81:844–52.

31. Land MF, Tatler BW. Steering with the head. The visual strategy of aracing driver. Curr Biol 2001;11:1215–20.

32. Imai T, Moore ST, Raphan T, Cohen B. Interaction of the body,head, and eyes during walking and turning. Exp Brain Res 2001;136:1–18.

33. Zikovitz DC, Harris LR. Head tilt during driving. Ergonomics 1999;42:740–6.

34. Friedman JM. Post-traumatic vertigo. Med Health R I 2004;87:296–300.

35. Johnson RT. Vestibular neuritis, or driving dizzily through Donegal.N Engl J Med 2004;351:322–3.

36. Lee H, Ahn BH, Baloh RW. Sudden deafness with vertigo as a solemanifestation of anterior inferior cerebellar artery infarction. J Neu-rol Sci 2004;222:105–7.

37. Black FO, Pesznecker S, Stallings V. Permanent gentamicin vestibu-lotoxicity. Otol Neurotol 2004;25:559–69.

38. Paige GD. Senescence of human visual–vestibular interactions. 1.Vestibulo-ocular reflex and adaptive plasticity with aging. J VestibRes 1992;2:133–51.

39. Cohen H, Heaton LG, Congdon SL, Jenkins HA. Changes in sensoryorganization test scores with age. Age Ageing 1996;25:39–44.

40. Nakayama M, Helfert RH, Konrad HR, Caspary DM. Scanningelectron microscopic evaluation of age-related changes in the rat ves-tibular epithelium. Otolaryngol Head Neck Surg 1994;111:799–806.

41. Rosenhall U. Degenerative patterns in the aging human vestibularneuro-epithelia. Acta Otolaryngol 1973;76:208–20.

42. Lyon MJ, Wanamaker HH. Blood flow and assessment of capillariesin the aging rat posterior canal crista. Hear Res 1993;67:157–65.

43. Lopez I, Honrubia V, Baloh RW. Aging and the human vestibularnucleus. J Vestib Res 1997;7:77–85.

44. Katsarkas A. Benign paroxysmal positional vertigo (BPPV): idio-pathic versus post-traumatic. Acta Otolaryngol 1999;119:745–9.

45. Cohen HS, Wells J, Kimball KT, Owsley C. Driving disability anddizziness. J Safety Res 2003;34:361–9.

46. Minor LB, Schessel DA, Carey JP. Meniere’s disease. Curr OpinNeurol 2004;17:9–16.

47. McKiernan D, Jonathan D. Driving and vertigo. Clin Otolaryngol2001;26:1–2.

48. Parnes LS, Sindwani R. Impact of vestibular disorders on fitness todrive: a census of the American Neurotology Society. Am J Otol1997;18:79–85.

49. Sindwani R, Parnes LS. Reporting of vestibular patients who are unfitto drive: survey of Canadian otolaryngologists. J Otolaryngol 1997;26:104–11.

50. Sindwani R, Parnes LS, Goebel JA, Cass SP. Approach to the vestib-ular patient and driving: a patient perspective. Otolaryngol HeadNeck Surg 1999;121:13–7.

51. Cohen HS. Vestibular disorders and impaired path integration alonga linear trajectory. J Vestib Res 2000;10:7–15.

52. Coeckelbergh TR, Brouwer WH, Cornelissen FW, Van Wolffelaar P,Kooijman AC. The effect of visual field defects on drivingperformance: a driving simulator study. Arch Ophthalmol 2002;120:1509–16.

53. Coeckelbergh TR, Cornelissen FW, Brouwer WH, Kooijman AC.The effect of visual field defects on eye movements and practicalfitness to drive. Vision Res 2002;42:669–77.

54. Ross JV. Bilateral homonymous hemianopsia caused by vascular oc-clusive disease. Am J Ophthalmol 1949;32:1588.–

55. Kearns TP, Wagener HP, Millikan CH. Bilateral homonymoushemianopsia; relationship to thrombosis of the basilar artery. AMAArch Ophthalmol 1954;53:560–5.

56. Reese FM. Bilateral homonymous hemianopsia. Am J Ophthalmol1954;38:44–57.

57. Zihl J, Wohlfarth-Englert A. The influence of visual field disorders onvisual identification tasks. Eur Arch Psychiatry Neurol Sci 1986;236:61–4.

58. Tant ML, Cornelissen FW, Kooijman AC, Brouwer WH. Hemi-anopic visual field defects elicit hemianopic scanning. Vision Res2002;42:1339–48.

59. Tant ML. Visual performance in homonymous hemianopia: assess-ment, training and driving [PhD Thesis]. Department of Psychology,



Division Biopsychology and Neuropsychology, University ofGroningen; 2002.

60. Moore ST, MacDougall H, Clark JB, Wuyts F, Lesceu X, Speyer JJ,Cohen B. Spatial disorientation—how the brain interprets linear ac-celeration during flight. J Vest Res 2004;14:114.

61. Gallimore JJ, Brannon NG, Patterson FR, Nalepka JP. Effects ofFOV and aircraft bank on pilot head movement and reversal errorsduring simulated flight. Aviat Space Environ Med 1999;70:1152–60.

62. Gallimore JJ, Patterson FR, Brannon NG, Nalepka JP. The opto-kinetic cervical reflex during formation flight. Aviat Space EnvironMed 2000;71:812–21.

63. Moore ST, Clement G, Raphan T, Cohen B. Ocular counterrollinginduced by centrifugation during orbital space flight. Exp Brain Res2001;137:323–35.

64. Mezey LE, Curthoys IS, Burgess AM, Goonetilleke SC, MacDougallHG. Changes in ocular torsion position produced by a single visualline rotating around the line of sight—visual ‘entrainment’ of oculartorsion. Vision Res 2004;44:397–406.

65. Angeli V, Benassi MG, Ladavas E. Recovery of oculo-motor bias inneglect patients after prism adaptation. Neuropsychologia 2004;42:1223–34.

66. Angeli V, Meneghello F, Mattioli F, Ladavas E. Mechanisms under-lying visuo-spatial amelioration of neglect after prism adaptation.Cortex 2004;40:155–6.

67. Wetzel PA, Anderson GM, Barelka BA. Instructor use of eye positionbased feedback for pilot training. In: Human Factors and ErgonomicsSociety 42nd Annual Meeting, Chicago, IL Oct. 5–9, 1998. Proceed-ings, vol 2. Santa Monica, CA: Human Factors and ErgonomicsSociety; 1998:abstract A99-1412602-54.

Steven T. Moore, PhDMt. Sinai School of Medicine

Neurology Dept., Box 11351 E 100th St., New York NY 10029

e-mail: [email protected]



functional assessment of head–eye coordination during ... · original article functional...

Documents