Norms and validity of the DriVR. a virtual reality driving assessment for persons with head injury

Transcription

1 Norms and validity of the DriVR a virtual reality driving assessment for persons with head injury Authors 1. Lili Liu, Ph.D., Associate Professor, Department of Occupational Therapy, Faculty of Rehabilitation Medicine, University of Alberta 2. Masako Miyazaki, M.Sc., Associate Professor, Director, Telehealth Technology Research Institute, University of Alberta 3. Benjamin Watson, Ph.D., Assistant Professor, Department of Computing Science, Faculty of Science, University of Alberta Source of work: University of Alberta Correspondence should be addressed to: Lili Liu, Ph.D., Associate Professor Department of Occupational Therapy, Faculty of Rehabilitation Medicine University of Alberta, 3-14 Corbett Hall Edmonton, Alberta, Canada. T6G 2G4 Phone: (403) Fax: (403) lili.liu@ualberta.ca 1

2 ABSTRACT This article presents the results of two studies of a new driving simulator called "DriVR". The first, normative study examined the performance of an uninjured driving population on the simulator. The driving ability of 148 participants in eight age categories was assessed using DriVR. The second, discriminative validity study examined DriVR s ability to discriminate between the performance of brain-injured and uninjured adults. We administered the DriVR assessment to 17 brain-injured adults. The performance of this group was compared to that of a subgroup of the uninjured participants, matched to the brain-injured participants in age, gender and years of education. In general, DriVR s measures showed age-related changes in participant performance, and were able to discriminate between head-injured and uninjured participants. These results suggested that age norms would be useful for analyzing the performance of headinjured clients who are being assessed with DriVR. It should be noted, however, that these studies did not examine DriVR s ability to predict performance in real, on-the-road tests. 2

3 INTRODUCTION According to the Medical Standards for Drivers used in Edmonton, Canada [1], individuals who have had a loss of consciousness associated with a head injury must receive a medical review to determine fitness to drive. This review consists of an assessment of cognitive, perceptual, motor and environmental factors [2]. Therapists, physicians or psychologists may provide clinical assessments of driving. These assessment results are then used to supplement an on-the-road test conducted by the Motor Vehicle Branch. Currently, there is no standardized clinical test that has been proven to be efficient and predictive of a brain-injured adult s fitness to drive. Professionals tend to use a battery of neuropsychological and cognitive perceptual tests, a driving simulator, and an on-the-road test [2, 3, 4, 5, 6]. This article describes the evaluation of the DriVR virtual reality (VR) simulator system for driving assessment. We begin with a review of existing driving assessment techniques and simulators. This is followed by a discussion of the advantages and disadvantages of VR simulators for driving assessment, and a review of related applications of VR in psychiatry and medicine. We then describe the DriVR system in detail, and present two studies evaluating DriVR: the first examined the performance of uninjured drivers on the system, and the second compared the performance of these drivers to a head-injured population. Michon s model of driving skills Regardless of the individual tests used, a common model followed by clinicians to categorize driving skills for assessment and intervention is the model developed by Michon [7, 8]. 3

4 Michon s model describes driving skills using three hierarchic levels. These are the Strategic, Tactical and Operational levels, from the highest to the lowest. Decisions and actions at a higher level require lower level skills. On the Strategic level, an individual makes decisions about the appropriate conditions for driving and plans how to arrive at a destination. Judgement, impulsive behavior and insight affect the quality of decisions and plans that a person makes at the Strategic level. To evaluate abilities at this level, clinicians can pose questions about an individual s metacognition. These questions require the client to demonstrate an awareness of his or her own abilities and limitations in driving.behaviors that affect Tactical decisions may the same as those at the Strategic level [8]. These behaviors include disinhibition, impulsiveness and abilities to plan and make good judgement. Behaviors that are specific to this level include driving at the appropriate speed in a residential neighbor hood versus a highway, turning on the headlights when visibility is reduced, deciding when it is safe to pass a car, and avoiding danger. These behaviors are frequently assessed using driving simulators. Simulators provide clients with a safe, yet dynamic situation in which the client can demonstrate his or her driving skills. On the Operational level, basic driving skills are examined. Visual scanning, spatial perception, lane tracking and operation of the steering wheel and brakes are all skills necessary for the operation of a motor vehicle. In addition, basic skills such as left-right discrimination, figureground perception and low reaction time are necessary for safe driving. Existing simulators for driving skill assessment Driving simulators are commonly used to assess skills at the Strategic, Tactical and Operational levels. Generally, clients are required to successfully complete a simulator test before taking an on-the-road test [2]. To date there have been a number of driving simulators 4

5 developed. These include the Doron driving simulator [4, 9, 10, 11], and the Computerized Driver Assessment Module (CDAM) [12, 13]. Simulators consist of the key components of a car such as a steering wheel, foot pedals and signal indicators. The client sits behind a screen on which a video or film presents driving scenes. Simulators typically provide scores on tactical and operational skills such as reaction time, lane tracking, signaling, ability to read and obey road and speed signs, threat recognition, and crash avoidance. The advantages of using simulators include: (a) objective scores of tactical and operational driving skills, (b) an opportunity to evaluate the client s strategic (metacognition and awareness) skills, (c) a safe environment for evaluators to observe a client s ability to drive in hazardous conditions, and (d) an opportunity for clients to practice driving and develop confidence before taking a road test. However, there are also limitations with using driving simulators. The validity of simulators has interested clinicians and researchers [5, 9, 10, 13]. In order for clients to drive the simulators as they would a real car, it is important that the simulators appear realistic to them [8]. Although most simulators present dynamic (moving) scenes of traffic, they are not interactive. This reduces the validity of the test. The association of simulator scores with expert judgements of actual driving in traffic has been reported to be poor by van Zomeren [8]. However, [10] reported that the Doron simulator, as part of a comprehensive evaluation, is capable of predicting behind-the-wheel performance. Cimolino and Balkovec [9] have found that the Doron Simulator is also useful as a basic evaluation tool for stroke clients, and a good training tool for adolescents. Recently, Richards [13] compared the performance of 32 uninjured participants with that of 50 head-injured participants on the CDAM (mean age in both groups was 34 years). The results show that discriminative validity of the CDAM is limited for brake reaction time, but not 5

6 for horn reaction time, visual reaction time, vehicle steering control, nor hazard avoidance. Another limitation with simulators is that they lack normative data [9]. Norms specific to a simulator help the clinician interpret a participant s performance relative to other individuals similar to his age, education or cognitive status. The use of VR in driving simulators VR offers a number of advantages over existing driving simulator technology. The most important of these is interactivity. In existing simulators, the displayed driving scene is a recording, and does not respond to driving control (e.g. turn, stop) from the driver. Clearly this back seat point of view is quite different from a real driving experience. VR is by definition an interactive technology, and responds to all control input from the driver. VR is also more immersive than existing driving simulators. Drivers assessed with existing simulators are able to see the room in which the simulator is located, particularly when they turn their heads. In fact, unlike real driving situations, turning the head when using existing simulators serves no purpose. VR typically makes use of a head-mounted display with head tracking technology, which not only obscures view of almost anything except the simulated view and the driving controls, but also allows the display to respond to head motion in a realistic manner. VR is also highly flexible. For the purposes of assessment, certain elements of the driving task and/or display might be changed in response to the driver s performance, possibly during the simulation itself. This would make adaptive assessment techniques possible. Updates in response to changes in automobile manufacture, traffic law, and cognitive or medical research should require only modification of existing simulation programs, rather than the filming of new simulation scenarios. Finally, recent developments in the computer industry have also made VR less expensive than existing 6

7 simulators. Rather than being centered around a high cost film or video projector, VR systems can now be run on low cost PC platforms. Of course, there are some potential disadvantages to VR-based driving assessment. VR systems make use of computer-generated graphics imagery, which appears less realistic than the video or film imagery used in existing driving simulators. The interactivity and immersiveness of VR may increase the likelihood of simulator sickness [14, 15, 16, 17], a nausea related to motion sickness. Nevertheless, neither of these shortcomings is severe enough to prevent use of VR, and as VR technology improves, the severity of these shortcomings will lessen. VR is already being successfully used for other psychiatric and medical applications. A summary of the research on psychotherapeutic applications of virtual reality technology can be found in [18]. A large portion this research has focussed on treatment of phobias [19, 20]. Researchers have demonstrated the effectiveness of VR for treatment of acro- and agoraphobia in controlled studies [21, 22]. Other reported research consists largely of promising pilot studies with only a few participants. Among these were studies investigating the application of VR for treatment of fear of flying [23], arachnophobia [24], and fear of public speaking [25]. Reports of research on therapeutic applications of VR not related to anxiety disorders are just beginning to appear. A study examining the use of VR for the treatment of eating disorders [26] treated 80 conference attendees and showed improvement of body-image ratings. Researchers examined the use of VR for teaching of children with learning disabilities in [27], and showed improvement of shopping skills in 19 children. In a related study [28], 18 learning disabled students being taught the Makaton sign language with a VR application showed improved three-dimensional learning. In [29], 24 students with movement disabilities showed the ability to locate objects on a map when given 3D virtual exposure to these locations. Reported 7

8 pilot studies have examined the application of VR to treatment of Parkinson s disease [30], training for wheelchair control [31], treatment of autism [32], and therapy for disturbed children [33]. Research investigating the use of VR for assessment (rather than therapy) is sparse. A survey of the issues expected to rise in the application of VR for assessment of cognitive and functional impairments may be found in [34]. A planned system for the assessment of body image is described in [35]. A successful case study describing the effectiveness of VR for cognitive assessment is described in [36]. Summary In summary, driving simulators have become an integral part of fitness-to-drive assessments in the clinic. They are used to evaluate strategic, tactical and operational driving skills. Simulators also provide a bridge between static tests and on-the-road tests. As researchers evaluate applications of VR for assessment, they will have to judge them by the same criteria used for all assessment tools: external validity (is the targeted skill truly being assessed), discriminative validity (can the system discriminate between highly and poorly skilled performance), and predictive validity (can the system predict real world performance). As VR simulators may not be realistic in every aspect, normative data is needed to establish a baseline with which to compare the performance of head-injured individuals. DESCRIPTION OF THE DriVR SIMULATOR 8

9 Hardware The DriVR created its imagery using a PC with a 166 MHz Pentium chip and 32 MB RAM. A graphics card provided 2D rendering acceleration. Experimental participants provided input to the system using a driving wheel and foot pedals for acceleration and braking from Thrustmaster. For visual display, DriVR used the Virtual I/O i-glasses. This head-mounted display provided a 30 degree horizontal field of view. The display was full color and stereoscopic, but was used in a biocular mode, with the same image sent to each eye. The i- glasses included three degree of freedom head tracker (yaw, pitch and roll). To reduce display jitter, changes in the roll of participants head position were ignored, and did not affect DriVR imagery. For audio display, DriVR used a Sound blaster card. Output from this card was sent to earphones in the head-mounted display. Content DriVR provided one practice and 10 testing scenarios, which appear in a continuous sequence as the participant drives through a small town roughly 1.4 km square. Table 1 summarizes the visual differences between each scenario, Figures 1 and 2 contain two example views. The motion of the vehicle being driven simulated automatic transmission, and was generated with a complex combination of variables including terrain, friction, and most importantly, the participants input from the steering wheel, brake and accelerator. Traffic signs and marks were modeled with textures on simple polygonal objects. Only the participants vehicle generated sound: a skidding sound, a collision sound, and an engine sound that varied with engine RPM, to give auditory feedback for speed. 9

10 Performance Mean frame rate was 14 Hz, and depending on the visual complexity of the current environment, varied between 12 and 18 Hz. Lag in head tracking was not measured but was quite apparent when participants turned their heads quickly. Lag in response to steering and brake/accelerator inputs was not appreciable in fact, some participants complained that response to this sort of input was too prompt, more prompt than in actual vehicles. Dependent Measures The DriVR system provided several measures to aid in assessment of tactical and operational skills we associate with driving. Some are used in several of the scenarios; others are used only in a specific scenario. Table 1 outlines the measures used in each scenario. Below we describe these measures. Speed: Measured the speeds of the vehicle in the scenario, in kilometers per hour. Mean, standard deviation, maximum and minimum statistics are provided. Lane: Measured the distance of the vehicle from the center of the appropriate lane in meters. Mean, standard deviation, maximum and minimum statistics are provided. Center: Counted the number of times the vehicle was steered across the road s center line. Shoulder: Counted the number of times the vehicle was steered off the edge of the road. Follow: Measured the distance of the vehicle from a car in front of it, in meters. Mean, standard deviation, maximum and minimum statistics are provided. Merge: Recorded whether or not the participant successfully steered the vehicle to merge with crossing traffic. Avoiding collision defined success. 10

11 Avoid: Recorded whether or not the participant successfully steered the vehicle to avoid a car backing out from a driveway. Avoiding collision defined success. Flag: Recorded whether or not the participant successfully steered the vehicle to comply with the directions of flag man holding a stop sign at a construction site. Stopping for at least three seconds defined success. Lane change: Recorded whether or not the participant successfully executed a lane change to avoid traffic cones closing the lane. Steering the vehicle to be in the correct lane when the cones were reached defined success. Driveway: Recorded whether or not the participant successfully executed a turn into a specific driveway. Stopping in the correct driveway defined success. Stop signs: Recorded the number of incorrect stops executed by the participant. This was a composite measure over all scenarios. EVALUATION METHODOLOGY We began with a pilot study designed to help us form a realistic experimental protocol. The nine participants in this study represented a wide range ages, from teenaged to elderly (mean 36.2 years). The results of this study are reflected in the experimental methodology described below. Each participant was given a pre-test questionnaire. The questionnaire contained items such as demographics (age, gender, years of education); history of head injury (date, duration of posttraumatic amnesia, history of seizures); list of medications or health conditions which may cause dizziness or nausea; and experience with driving (years with a valid driver s license and years of driving experience). 11

12 DriVR contained a familiarization route, designed to allow participants to become comfortable with the VR display and controls. Based on the results of the pilot study, we allowed each subject a maximum of 10 minutes of familiarization time. After a break, subjects completed two runs (trials) through the complete sequence of DriVR scenarios. Using two runs allowed us to measure any effect of practice. To avoid fatigue, participants were required to take a fiveminute break between the two trials. Participants were able to complete familiarization and two trials in well under an hour. Throughout testing, the evaluator carefully observed each participant for signs of dizziness or nausea. If a participant was unable to complete the session, his or her data was discarded and another participant was recruited to replace the discarded data. Upon completion of testing, participants rated the perceived difficulty of the DriVR assessment on a visual analogue scale of 0 to 10, with 0 indicating least difficulty. The research proposal received ethical approval from the University of Alberta. All participants were required to sign an informed consent form. Drivers less than the age of 18 years were required to have a parent or guardian sign the consent form. We were aware that some participants, particularly the head-injured adults, may have been concerned about potential uses of their test results. All participants were reassured that the predictive validity of the DriVR had not been determined, and that this was not the purpose of this study. They were informed that, as volunteers, they were assisting us with the evaluation of a system for potential use as a driving simulator. TYPICAL PERFORMANCE WITH DriVR 12

13 Objective The objective of this study was to find how a typical driving population would perform on the DriVR simulator. The results of this study could then be used to evaluate the external validity of the DriVR measures, and in comparison with the performance of an atypical (e.g. head-injured) driving population. Participants Normative data was collected using a convenience sample of young, middle aged and older adults living in the city of Edmonton. Our goal was to test twenty participants in each of the following eight age categories: less than 16, 16-25, 26-35, 36-45, 46-55, 56-65, 66-75, 76 and older. We chose to include a group of participants younger than the legal driving age of 16 years because we were interested in how this group, whom we assumed were familiar with computer game simulations, would perform relative to the older groups. We attempted to have an equal distribution of men and women in each age group. No participant had a history of head injury. Participants were recruited through word of mouth and advertisements placed in newspapers targeted at University students, the general public and seniors. Figure 3 shows the number of participants tested by age group. From a total of 162 participants, 148 completed DriVR testing (73 men and 75 women). 14 participants were unable to complete the assessment due to nausea or physical discomfort. With the exception of the group over the age of 75, 20 participants were tested in each group. Due to the severe winter weather in our city, few seniors over the age of 75 could travel to the test site independently. Figure 4 depicts the average number of years of education by age group. The only significant trend is the understandably lesser degree of education in the youngest age group. 13

14 Figure 5 depicts driving history by age group. The number of years participants had been driving and in possession of a driver s license increased with age. Driving frequency was roughly equivalent across all groups (excepting again the youngest group). Participants took an average 14.9 minutes to complete the test (SD=15.1, range: minutes). Figure 6 shows that completion time increased with age. The correlation between time and age was statistically significant (Pearson s r =.56, p<.01). Dependent measures Driving speed. Speed limits varied across and within scenarios, but was most often 40 or 50 km/hr. Figure 7 shows the average speed of each age group for five of the DriVR scenarios. Average speed to decreased with age. All of the average speeds were at or below the posted speed limits. Lane and follow. Figure 8 shows the average distance from the center line by age group for each of the curved road, sloped road, shop road and opposite traffic scenarios. The average distances for each scenario were 2.1, 2.3, 2.1 and 2.3 meters. The average distance by age group from the lead car in the follow traffic scenario is depicted in Figure 9. Participants in the three older age categories kept a further distance from the lead car due to their slower speeds. Avoid, lane change, merge and driveway. Figure 10 describes the percentage of uninjured participants by age group who passed traffic merge, traffic avoidance, lane change and driveway selection. While traffic avoidance and driveway selection were successfully performed by the majority of the participants, close to half of the participants failed the traffic merge and lane change requirements, suggesting that performance on these two latter tasks are not good indicators of uninjured performance. 14

15 Stop signs, shoulder and center. The percentage of participants who correctly stopped for all stop signs declined with age (see Figure 11). It is interesting to note that the only group in which no participant stopped correctly for all stop signs was the age group, in which participants were at the pre-driving age. The percentage of participants who did not cross the shoulder line or centre line are depicted in Figures 12 and 13 respectively. Perceived difficulty. Figure 14 shows the perceived difficulty by age group. The average was 6.5 (SD=1.9). Results in the flag man scenario were not meaningful in either of the studies presented in this article. For reasons of brevity they are not discussed. Discussion A number of measures indicated that there were age-related differences in performance on the DriVR. Driving speed decreased with age, while completion time increased with age. These speed differences provided an explanation for the increase in following distance with age. The pre-driving age group was particularly poor by the avoid, lane change and driveway measures, and was. the only group in which no participant stopped correctly for all of the stop signs. This might be explained by this group s tendency to drive faster. Did DriVR measure real driving ability? These results raised doubts. Some of the continuous measures showed results inconsistent with real world performance. For example, average driving speed seemed unusually low. Many of the success/failure measures showed results that would be disastrous in the real world. In a particularly alarming example, only 30% of drivers between the ages of 36 and 45 correctly stopped at all stop signs (see again Figure 11). 15

16 Based on these results, we recommended that the DriVR assessment be simplified and adjusted so that it would yield results one would expect in the real world. However, it could be argued that the usefulness of a driving simulator lies primarily in its ability to predict performance on a real world driving assessment (predictive validity), not in the similarity of its assessment results to real-world results (external validity). Further research should directly evaluate the predictive validity of the DriVR system. SENSITIVITY OF DriVR TO PERFORMANCE OF HEAD-INJURED DRIVERS Objective The objective of this study was to determine whether or not the DriVR simulator could discriminate uninjured drivers from drivers that had previously suffered head-injury. This was one of the original design goals of the DriVR system. Participants Discriminative validity of the DriVR was evaluated using a group of 17 head-injured adults. The performance of these participants was compared to a group of 17 uninjured participants from the first, normative study. The uninjured group was matched in gender, age and education to the head-injured group. All head-injured participants resided in the community at the time of testing, and had had a valid driver s license before their injury occurred (they did not necessarily hold a valid driver s license at the time they participated in the study). Head-injured participants were recruited from the community through advertisements. 16

17 15 head-injured participants were completely assessed, two could not complete the testing due to discomfort. Thirteen of these participants were men and two were women. The average ages for the head-injured and uninjured groups were 40.3 (SD = 18.5) and 40.0 (SD = 17.9) years, respectively. Head-injured participants had and average of 13.7 (SD = 1.9) years of education, while the uninjured group had 14.8 (SD = 2.6) years of education. Six head-injured participants did not own drivers licences and an additional five were not driving. Four had restricted drivers licences. All of the uninjured participants owned licences and only one was not driving. Head-injured participants had had their drivers licences for a mean of 16.3 (SD = 19.2) years, while uninjured participants had had their licences for an average of 22.6 years (SD = 17.5). This difference did not reach statistical significance (t(df) = -.9 (28), p=.36). On average, the head-injured participants drove on 3.8 days each week, compared to 5.4 days for the uninjured participants. Again, this difference was not statistically significant (t(df) = - 1.5(28), p=.14). Head-injured and uninjured participants completed the test in an average of 14.4 and 13.4 minutes, respectively. Dependent measures Driveway, merge, lane change and avoid. The percentages of participants succeeding by these measures in each group are shown in Figure 15. Uninjured participants were more successful in all but traffic avoidance. Center and shoulder. The percentages by group of participants who did not cross the center or shoulder lines are shown in Figures 16 and 17. Head-injured participants crossed the center and shoulder lines much more often than uninjured participants. 17

18 Stop signs. Figure 18 depicts the percentage by group of participants who made a specified number of incorrect stops. The percentage of head-injured participants making a specified number of incorrect stops always exceeded the matching uninjured percentage. Only head-injured participants stopped incorrectly at four stop signs, while only uninjured participants were able to stop correctly at all stop signs. Perceived difficulty. When asked to rate their perceived difficulty with DriVR, head-injured participants gave a mean rating of 6.6 (SD=2.4), while uninjured participants provided a mean rating of 5.5 (SD=2.5). The difference between the groups was not statistically significant (t(df)=1.2(28), p=.23). Discussion DriVR was able to discriminate between the performance of head-injured and uninjured participants. Given that this was a goal of DriVR s design, this is quite encouraging. Nevertheless, it is important to consider other possible explanations for the differences in measured performance between the two groups. For instance, head-injured participants had had their driver s licenses an average of five years less than uninjured participants, and drove less frequently than uninjured participants. These differences were not statistically significant, and it is not certain how they may have affected measured results. The data we collected did not reveal whether these differences were due to the head injury or other factors. GENERAL DISCUSSION The main purpose of this study was to examine the feasibility of using a new driving simulator for evaluating fitness-to-drive in individuals who have sustained a head injury, and are 18

19 still living in the community. This new evaluation tool, called DriVR, used virtual reality (VR) to simulate driving scenarios that we assumed tested the Strategic, Tactical and Operational skills required for on-the-road driving. As this was a new tool, we needed to examine how normal, uninjured people perform. We also needed to study the DriVR tool s ability to discriminate between people who had a head-injury from the uninjured population. The DriVR tool was able to discriminate between head-injured and uninjured participants by most measures. However, the ability of DriVR to predict performance in real, on-the-road tests must still be evaluated. The results of this study also have wider implications for the use of VR for driving assessment in particular, and other sorts of skill assessment in general. These implications are given special weight by the large size of the population used in the normative study. The advantages of VR s interactivity are fairly clear. Driving, like most human activities, is an interactive experience every action has an effect, and these effects in turn inform the driver s next action. In traditional simulators, this feedback loop is broken the motion of the displayed car has no correspondence to the driver s action on the steering wheel, accelerator pedal, or brake pedal. In a VR simulator, drivers can progress at different speeds and follow different paths. This gives added meaning to the measures (e.g. speed, following distance) made by a VR simulator. Moreover, since the driver has this control, repetitive drilling or testing (like that used here) with VR simulators is useful, while this sort of repetition quickly becomes pointless with traditional simulators. VR s interactivity also has its disadvantages: reduced visual fidelity, and the nausea that interactivity and immersiveness can bring. The visual display used in this study was certainly not realistic, but measures were still able to discriminate between head-injured and uninjured 19

20 participants. This bodes well for the application of VR to assessment. Almost 10% of 185 participants were unable to complete the study due to nausea. Other participants experienced milder symptoms that may have affected their performance. Future VR simulators will have to attempt to reduce these effects, without driving up simulator costs. The elderly have largely been ignored in the design of existing VR applications. This must change as VR is increasingly applied in the medical and psychiatric fields, in which care for the elderly is so important. This evaluation of a VR application is one of the few to date that has included a significant number of the elderly. We note two trends in our results that show promise for VR applications for the elderly. First, despite being largely unfamiliar with electronic games and computers, elderly participants were able to complete testing with performance quite comparable to participants in other age groups. Second, elderly participants were not more susceptible to the nausea of simulator sickness than other age groups. ACKNOWLEDGMENTS We thank Lisa Kovacs for assisting us with data collection and data entry. This study was funded by a grant from Imago Systems, Inc., Vancouver, British Columbia. The preliminary results of this study were presented at the annual conference of the Canadian Association of Occupational Therapists in Halifax, Nova Scotia, June Lili Liu, PhD, Assoc. Prof., (403)

21 REFERENCES 1. Transportation and Utilities (1994). Medical standards for drivers. Edmonton, AB: author. 2. Katz, R. T., Golden, R. S., Butter, J., Tepper, D., Rothke, S., Holmes, J., & Sahgal, V. (1990). Driving safety after brain damage: follow-up of twenty-two patients with matched controls. Archives of Physical Medicine and Rehabilitation, 71, Engum, E. S., Lambert, E. W., & Scott, K. (1990). Criterion-related validity of the cognitive behavioral driver s inventory: brain-injured patients versus normal controls. Cognitive Rehabilitation, 8, Galski, T., Bruno, R. L., & Ehle, H. T. (1993). Prediction of behind-the-wheel driving performance in patients with cerebral brain damage: a discriminant function analysis. The American Journal of Occupational Therapy, 47, Handler, B. S., & Patterson, J. B. (1995). Driving after brain injury. Journal of Rehabilitation, 61, Korner-Bitensky, N., Sofer, S., Kaizer, F., Gelinas, I., & Talbot, L. (1994). Assessing ability to drive following an acute neurological event: are we on the right road? Canadian Journal of Occupational Therapy, 61, Michon, J.A. (1979). Dealing with danger. Summary report of a workshop in the Traffic Research centre, State University, Groningen. 8. van Zomeren, A. H., Brouwer, W. H., & Minderhoud, J. M. (1987). Acquired brain damage and driving: a review. Archives of Physical Medicine and Rehabilitation, 68, Cimolino, N. & Balkovec, D. (1988). The contribution of a driving simulator in the driving evaluation of stroke and disabled adolescent clients. Canadian Journal of Occupational Therapy, 55,

22 10. Galski, T., Bruno, R. L., & Ehle, H. T. (1992). Driving after cerebral damage: a model with implications for evaluation. The American Journal of Occupational Therapy, 46, Quigley, F.L., & DeLisa, J.A. (1983). Assessing the driving potential of cerebral vascular patients. American Journal of Occupational Therapy, 37, Bowen, LA, Hershler, C, Laszlo, C, & Miller, F (1989, June). Use of the computerized driver assessment module (CDAM) for the assessment of the brain injured individual. Proceedings of the 12th annual conference of (RESNA) Rehabilitation Engineering Society of North America, (now called the Association for the Advancement of Rehabilitation Technology), Richards, G., Petrov, T., & Laszlo, C (1995). Driving assessment of the cognitively impaired: preliminary report on the CDAM III Driving Simulator. 14. Kennedy, R., Lane, N., Lilienthal, M., Berbaum, K. & Hettinger, L. (1992). Profile analysis of simulator sickness systems: applications to virtual reality systems. Presence, 1, 3, Kennedy, R. & Stanney, K. (1997). Postural instability induced by virtual reality exposure: development of a certification protocol. International Journal of Human-Computer Interaction, 8, 1, Stanney, K. & Kennedy, R. (1997). Development and testing of a measure of the kinesthetic position sense used to assess the aftereffects from virtual environment exposure. Proc. IEEE Virtual Reality Annual International Symposium (VRAIS), Los Alamitos, CA, USA: IEEE Computer Society Press. 17. Stanney, K. & Kennedy, R. (1997). The psychometrics of cybersickness. Communications of the ACM, 40, 8,

23 18. Wiederhold, B.K. & Wiederhold, M.D. (1998). A review of virtual reality as a psychotherapeutic tool. CyberPsychology & Behavior, 1, 1, Glantz, K., Durlach, N.I., Barnett, R.C. & Aviles, W.A. (1997). Virtual reality (VR) and psychotherapy: opportunities and challenges. Presence, 6, 1, Strickland, D., Hodges, L., North, M. & Weghorst, S. (1997). Overcoming phobias by virtual exposure. Communications of the ACM, 40, 8, Hodges, L., Rothbaum, B., Kooper, R., Opdyke, D., Meyer, T., North, M., de Graff, J., & Williford, J. (1995). Virtual environments for treating fear of heights. IEEE Computer, 28, 7, North, M., North, S. & Coble, J. (1996). Effectiveness of virtual environment desensitization in the treatment of agoraphobia. Presence, 5, 3, Hodges, L., Rothbaum, B., Watson, B., Kessler, D. & Opdyke, D. (1996). Virtually conquering fear of flying. IEEE Computer Graphics & Applications, 16, 6, Carlin, A. Hoffman, H. & Weghorst, S. (1997). Virtual reality and tactile augmentation in the treatment of spider phobia: a case report. Behavioral Research Therapy, 35, 2, Virtual reality combats fear of public speaking (1997). Real Time Graphics, July, S Riva, G. (1996). The virtual environment for body image modification (VEBIM): development and preliminary evaluation. Presence, 6, 1, Brown, D., Kerr, S. & Wilson, J. (1997). Virtual environments in special-needs education. Communications of the ACM, 40, 8, Standen, P. & Low, H. (1996). Do virtual environments promote self directed activity? A study of students with severe learning difficulties learning Makaton sign language. Proc. 1st 23

24 European Conference on Disability, VR and Associated Technology, Maidenhead, UK: ECVRAT. 29. Foreman, N., Wilson, P. & Stanton, D. (1997). VR and spatial awareness in disabled children. Communications of the ACM, 40, 8, Riess, T. & Weghorst, S. (1995). Augmented reality in the treatment of Parkinson s disease. Proc. Medicine Meets Virtual Reality III, Bowman, T. (1997). VR meets physical therapy. Communications of the ACM, 40, 8, Strickland, D. (1996). A virtual-reality application with autistic children. Presence, 5, 3, Kijima, R., Shirakawa, K., Hirose, M. & Nihei, K. (1994). Virtual sand box: development of an application of virtual environments for clinical medicine. Presence, 3, 1, Rizzo, A., Buckwalter, J., Neumann, U., Kesselman, C. & Thiebaux, M. (1998). Basic issues in the application of virtual reality for the assessment and rehabilitation of cognitive impairments and functional disabilities. CyberPsychology & Behavior, 1, 1, Riva, G. (1998). Virtual reality in psychological assessment: the body image virtual reality scale. CyberPsychology & Behavior, 1, 1, Mendozzi, L., Motta, A., Barbieri, E., Alpini, D. & Pugnetti, L. (1998). The application of virtual reality to document coping deficits after a stroke: report of a case. CyberPsychology & Behavior, 1, 1,

25 Table 1. Descriptions of the 10 scenarios in the order they appeared in the DrivVR system. Scenario Traffic Signs Other Objects Moving Cars Measure Name Spd Stp Oth Bldg Oth Num Flow Types Curved Road 2 12 lamp speed, lane, center, shldr Sloped Road opp speed, lane, center, shldr Follow Traffic mark 12 same, opp speed, flw, center, shldr Shop Road pk cr 2 same speed, lane, center, shldr Traffic Merge 1 1 turn 5 cross merge Avoid Traffic cross avoid Opposite Traffic tunnel 8 opp speed, lane, center, shldr Flag Man 1 ped 2 same flag Lane Change cone 9 5 opp lane ch Driveway Choice 4 1 pk cr 4 drwy drwy Note. Spd = speed limit signs, Stp = stop signs, turn = turn arrow signs, cone = traffic cones. Bldg = buildings, lamp = lamposts, mark = road markers, pk cr = parked cars, ped = pedestrian, drwy = driveway. Same = same as vehicle, opp = opposite. Shldr = shoulder, flw = follow, lane ch = lane change. 25

26 Figure 1: Entering the sloped road scenario. Notice the traffic signs, the other vehicle, and the terrain in the background. Figure 2: The traffic merge scenario, with speed limit sign. Figure 3: Number of participants who completed (or could not complete) DriVR by age group. Figure 4: Years of education of participants by age group. Figure 5: Driving history means by age group. Figure 6: Mean time in minutes to testing completion by age group. Figure 7: Average speed in five scenarios by age group. Figure 8: Average distance from the center line in four DriVR scenarios, by age group. Figure 9: Average distance from the preceding car by age group. Figure 10: Percentage of participants who passed the avoid, lane change, merge and driveway measures by age group. Figure 11: Percentage of participants who stopped correctly at all stop signs by age group. Figure 12: Percentage of participants who did not cross the shoulder line in five scenarios, by age group. Figure 13: Percentage of participants who did not cross the center line in five scenarios, by age group. Figure 14: Subjective difficulty ratings by age group. 0 indicates least difficulty. Figure 15: Percentage of head-injured and uninjured participants who passed the avoid, lane change, merge and driveway measures. Figure 16: Percentage of head-injured and uninjured participants who did not cross the center line in five scenarios. 26

27 Figure 17: Percentage of head-injured and uninjured participants who did not cross the shoulder line in five scenarios. Figure 18: Percentage of head-injured and uninjured participants who made 0, 1, 2, 3 or 4 incorrect stops. 27

28

29

30 Not completed Completed

31

32 yrs w/ license yrs driving days driving/wk

33

34 curved road sloped road follow traffic shop road opposite traffic

35 curved road sloped road shop road opposite traffic

36

37 % pass avoid % pass lane chg % pass merge % pass driveway

38

39 curved road sloped road follow traffic shop road opposite traffic

40 curved road sloped road follow traffic shop road opposite traffic

41

42 driveway merge lane change Uninjured Head-injured avoid

43 opposite traffic shop road follow traffic Uninjured Head-injured sloped road curved road

44 opposite traffic shop road follow traffic Uninjured Head-injured sloped road curved road

45 4 incorrect stops 3 incorrect stops 2 incorrect stops Uninjured Head-injured 1 incorrect stop 0 incorrect stop