Abstract
Purpose.:
To compare eye and head movements, lane keeping, and vehicle control of drivers with hemianopic and quadrantanopic field defects with controls, and to identify differences in these parameters between hemianopic and quadrantanopic drivers rated safe to drive by a clinical driving rehabilitation specialist compared with those rated as unsafe.
Methods.:
Eye and head movements and lane keeping were rated in 22 persons with homonymous hemianopic defects and 8 with quadrantanopic defects (mean age, 53 years) who were ≥6 months post-injury and 30 persons with normal fields (mean age, 53 years). All were licensed to drive and were current drivers or aimed to resume driving. Participants drove a 6.3-mile route along non-interstate city roads under in-traffic conditions. Vehicle control was assessed objectively by vehicle instrumentation for speed, braking, acceleration, and cornering.
Results.:
As a group, drivers with hemianopic or quadrantanopic defects drove slower, exhibited less excessive cornering or acceleration, and executed more shoulder movements than the controls. Those drivers with hemianopic or quadrantanopic defects rated as safe also made more head movements into their blind field, received superior ratings regarding eye movement extent and lane position stability, and exhibited less sudden braking and drove faster than those rated unsafe.
Conclusions.:
Persons with hemianopic and quadrantanopic defects rated as safe to drive compensated by making more head movements into their blind field, combined with more stable lane keeping and less sudden braking. Future research should evaluate whether these characteristics could be trained in rehabilitation programs aimed at improving driving safety in this population.
There has been considerable debate in the literature regarding the driving safety of individuals with homonymous hemianopic and quadrantanopic field defects.
1 –5 An important consideration for understanding their driving performance is the extent to which individuals with these field defects might adopt patterns of eye and head movements that assist them to compensate for their field loss. If this were the case, it would provide justification for exploring the potential for predicting whether an individual with these field defects might have the potential for safe driving and for training scanning behaviors as a means of improving driver safety in these individuals.
6 –10
Numerous studies have explored the eye and head movements and scanning behavior of persons with homonymous hemianopia in well-controlled laboratory-based settings; however, none have been conducted under real-world driving conditions. These laboratory-based studies have shown that persons with hemianopic field defects mainly look toward their blind hemifield when undertaking a range of tasks, including counting dots,
11 –13 viewing natural and degraded images,
14 viewing randomly presented
13 and moving targets
15 within a virtual reality environment, but not when assembling wooden models in a static environment.
11 Martin et al.
11 explain their findings by suggesting that these compensatory strategies of biasing gaze in the direction of the blind hemifield are most evident in dynamic and unpredictable environments, where subjects cannot rely on their spatial memory to locate salient objects. This hypothesis was recently supported by Hardiess et al.,
13 who found that the differences in gaze patterns between hemianopes who performed visual search tasks at “adequate” or “inadequate” levels were dependent on the level of complexity of the search task. They suggested that the poorer performance of the inadequate performers on the more complex task was due to reduced working memory. Given that driving is a complex and dynamic task, where drivers cannot rely on their spatial memory to locate salient objects, we hypothesize that individuals with hemianopic defects might similarly adopt head and eye movements that bias fixation toward the blind field while driving, and that those who adopt these strategies will be able to successfully compensate for their field defects and exhibit safer driving performance.
In addition, few studies have assessed the on-road driving characteristics of hemianopic and quadrantanopic drivers, including speed, braking, acceleration, cornering, and lane keeping, which might also differentiate between safe and unsafe drivers. Szlyk et al.
1 in an interactive driving simulator study reported higher numbers of lane boundary crossings for a small sample of persons with hemianopia compared with controls, while on-road studies have also reported problems with steering stability and lane keeping.
2,5 Bowers et al.
16 in a driving simulator study also showed that hemianopic persons adopted a lane position toward their seeing field, therefore providing a safety margin on their blind side. However, this finding has not been verified for actual on-road driving performance, an issue that is addressed in this study.
The aim of the present study was thus to compare the patterns of eye and head movements, lane keeping, and vehicle control of drivers with homonymous hemianopia and quadrantanopia to that of age-matched drivers with normal visual fields while driving under real-world conditions. We also compared the eye and head movements of those hemianopic and quadrantanic drivers rated as safe to drive with those rated as unsafe. We hypothesized, based on the evidence of previous studies, that persons with hemianopia would make more head movements into their blind field as a means of compensating for their field defects and that this would be more evident in those rated as safe to drive. We also hypothesized that those rated as unsafe to drive would adopt a lane position in the direction of their seeing field to avoid their blind side, while those rated as safe to drive would maintain a relatively central lane position.
On-road driving performance was assessed under in-traffic conditions in an automatic transmission vehicle (Chevrolet Impala 2007), instrumented to measure acceleration and deceleration, lateral/longitudinal forces, vehicle speed, and recording of the internal and external driving environment (Vigil Vanguard System, Brisbane, Australia). An accelerometer was mounted on the roof of the vehicle along with inertial sensors to record braking and acceleration forces, while a roof-mounted GPS system sampled the speed and position of the vehicle. Three cameras were mounted on the roof of the vehicle (one each to the extreme left and right of the vehicle and pointing slightly downwards to record the position of the vehicle front right and left fenders for assessment of lane position), and one mounted in the center of the vehicle to record the forward road scene.
An internally mounted camera pointing directly toward the participant's face and upper torso was used to record the pattern of head and eye movements from which an index of eye and head movements was derived post-testing. Although this does not provide a quantitative analysis of fixation durations, saccades, and head movements, it provides a good basis for identifying and further exploring any differences in eye and head movement patterns between drivers with hemianopic or quadrantanopic field defects and those with normal fields. This was necessary because recording eye movements in the field, under ever-changing outdoors conditions while the participant is actually driving, is much more challenging than in a laboratory setting or driving simulators where there is excellent level of control of lighting and participant location relative to the scene ahead.
The driving performance of each participant was assessed under in-traffic conditions along 6.3 miles of non-interstate driving in residential and commercial areas of a city as described previously.
5,18 Drives were held between 9 AM and 3 PM to avoid rush hour traffic and were cancelled if it was raining or the road was wet. A certified driving rehabilitation specialist (CDRS) who was also a licensed occupational therapist sat in the front passenger seat of the vehicle; she has eight years of clinical experience in driving assessment and rehabilitation of patients with a wide variety of medical and neurologic conditions. The CDRS evaluated driving performance, had access to a dual brake, and was responsible for monitoring safety and was aware of the medical and functional characteristics of the participants she was evaluating on the road, as is standard practice. However, because of the potential for bias and its impact on interpreting the results, we were also interested in the extent to which her ratings of driving safety agreed with two backseat raters who ensured appropriate operation of the vehicle's instrumentation and recording system throughout the drive and were completely masked to the visual field (i.e., hemianopia/quadrantanopia/normal) and health characteristics of each participant.
Each drive began by participants completing a series of basic driving maneuvers in a parking lot to ensure they had adequate vehicle control and to become familiar with the vehicle. Once the participant exhibited adequate vehicle control, the on-road driving evaluation began, starting in quiet city streets in a residential neighborhood and then proceeding to busier roads. The CDRS used a five-point rating system to assess different components of driving performance, as well as to derive an overall rating of performance: 1 = driving was so unsafe that the drive was terminated; 2 = exhibited a couple of unsafe maneuvers but did not reach the level of drive termination; 3 = driving was unsatisfactory but not unsafe at that time given the traffic circumstances; 4 = driver exhibited a few minor driving errors; and 5 = there were no obvious driving errors.
18 Scores of 1 and 2 were classified as failing the driving assessment and being unsafe to drive, whereas scores of 3, 4, and 5 were considered to be passes.
There was perfect agreement between the CDRS and the backseat evaluator in terms of determining which of the drivers passed or failed the driving assessment,
18 which provides important validation regarding the reliability of the CDRS's judgments with respect to safe driving (the study's main dependent variable).
The data collected by the instrumented vehicle were exported as text and graphical files and examined using specialized software (Vigil Vanguard System), which automatically generated outcome scores of driving speed and excessive force events defined as jerky cornering, sudden braking, and acceleration. Excessive or jerky acceleration, braking, and cornering were defined as when >0.2g was exerted and recorded by the detecting sensors. This value was set as the default by the system, being defined as the force level that typically feels uncomfortable to a passenger riding in a vehicle.
The videos of the external environment were analyzed to rate road position and those of the internal vehicle environment to count head movements and rate eye movements. Two independent research assistants who were completely masked to the visual field (i.e., hemianopia/quadrantanopia/normal), and health characteristics of each participant, or their driving category (safe/unsafe), conducted an analysis of the driving videos using a scoring system that allowed quantitative scoring of head movements, categorization of the extent of eye movements, and rating of lane positioning (given that the lane markings along the route were clearly evident only for some sections of the driving route). Sideways head movements were categorized into small and large head movements, where small head movements were defined as movements ranging from the forward-facing position to a 45° angle (selected as the halfway position between a forward facing position and one where the driver was looking directly sideways at 90° to the camera view), with large head movements defined as those that were greater than 45°. Head movements that were around the borderline 45° position, where it was difficult to determine whether they fell into the small or large categories, were always classified as small for the purposes of consistency. Counts were then made of each movement by category and direction (left or right). Vertical head movements were counted, as were shoulder movements.
A five-point Likert-type scale was used to categorize eye movements (1 = few saccades, 3 = average number of saccades, 5 = many saccades), road position (1 = very poor/unstable, 3 = some errors, 5 = very good/stable and whether a central position or to the left or the right of the lane for the majority of the drive), and head movements overall (1 = not excursive, 3 = some excursions, 5 = highly excursive). These categorizations were made once the raters had observed all the videos and so had a clear impression of the range of performance across all participants.
The intra-rater reliability of the two research assistants scoring the driving videos ranged from r = 0.51, P < 0.0001, for small right head movements to r = 0.85; P < 0.0001, for the total number of head movements. The data from the two raters were thus combined to provide an average rating for all the head movement counts and overall scoring of head and eye movements and lane position. Driving videos from a random sample of nine participants were reanalyzed by the research assistants to derive a measure of their own scoring repeatability that ranged from intraclass correlation coefficient (ICC) = 0.65, P < 0.0001, to ICC = 0.97, P < 0.0001.