Virtual Reality-Base Intelligent Orientation and Mobility Specialists Trial

The human subject research is a randomized, controlled training trial that tests the effectiveness of three Virtual Reality-based Intelligent Orientation and Mobility Specialists (VR-IOMSs) in teaching orientation and mobility (O\&M) task skills to individuals with low vision (LV). It will be conducted on two sites, University of Alabama at Birmingham (UAB) and Alabama Institute for Deaf and Blind (AIDB). The same protocol will be used on both sites. UAB will be the sIRB for the trial.

Three O\&M tasks will be studied, timing to cross a signalized street using the near lane parallel traffic surge skill (Task NLPT), timing to cross an uncontrolled street using the traffic gap judgment skill (Task TGJ) and learning outdoor numbering system (Task ONS). A VR-IOMS will be developed and validated for each task. The training does not involve research subjects walking into street traffic.

Low vision subjects who have difficulties with these O\&M tasks due to their impaired vision will be recruited from UAB and AIDB according to procedures approved by UAB IRB. Specifically, clinicians of UAB Center for Low Vision Rehabilitation will identify potential study subjects from records of existing patients and from new patient population. Recruitment letters from the clinicians will be sent to potential subjects. Patients who are interested in participating will initiate contact with UAB research team for initial evaluation and scheduling an enrollment visit. Certified O\&M specialist (COMS) faculty of AIDB's School for the Blind, the Gentry Facility and Alabama Freedom Center for the Blind will identify potential study subjects from new students/clients and will use a recruitment script to inform them about the study. Students/clients who are interested will contact the AIDB team to arrange initial evaluation and enrollment session.

Initial evaluation includes inquiries of visual and hearing impairment, mental status and O\&M training history. Potential subjects who pass the initial evaluation will attend an enrollment session where demographic information will be collect, visual and hearing functions will be assessed, if necessary and functional vision and ability to perform the study O\&M tasks will be evaluated by COMS on real streets. Informed consent/assent material approved by UAB IRB will be presented to potential subjects who meet the inclusion criteria and/or their legally authorized representatives. Research activity will not begin until written informed consent/assent is obtained according to approved procedure.

Enrolled subjects will be randomized into three groups, learning the task from a VR-IOMS (experimental group), from a human COMS in real streets (active control group) and not learning the task but spending the same amount of time watching LV education videos (placebo group). All subjects will be evaluated by COMSs in real streets around the two study sites before training (pre-training), within 3 days after the completion of training (post-training) and 3 months after the completion of training (follow up). Their ability to perform the O\&M tasks in real streets will be assessed quantitatively using objective methods. Specifically, in Task NLPT, the performance is measured by a safety score, which is determined by the time of the onset of the steady pedestrian WALK signal, the time when the first straight going car in the same direction as the traveler's crossing direction entering the street intersection and the time when the subject decides to start crossing (say GO). These times will be measured by the COMS using a stopwatch. The safety score, ranges from 0 to 1, represents the proportion of the duration of the pedestrian phase after the subject says GO. A larger score indicate a larger portion of pedestrian phase is available for the subject to walk across the street and is thus safer. A GO call in the steady DON'T WALK phase will be recorded as a negative number and is unsafe. This safety score was developed and used in the investigators' pilot research. The rule of thumb used by COMSs, start crossing within 7 seconds after the onset of the pedestrian WALK signal, will also be used. In Task TGJ, the performance is measured by a safety margin, which is the traffic gap width in seconds (the time interval in seconds between the first and second car passing in front of the traveler) minus the delay of the subject's crossing decision (say GO) and minus the time needed for the subject to walk across the street at the participant's preferred walking speed. Validated objective methods to measure the three times involved in safety margin determination have been used in previous studies of low vision patient TGJ performance and will be used in this research. A positive safety margin represents the time left of the traffic gap when the subject has reached the opposite side of the road, thus indicates a safe crossing decision. A negative safety margin represents the time when the subject is still in the road after the traffic gap has closed. It indicates an exposure of the subject to the traffic and thus indicates an unsafe decision. In Task ONS, the performance is measured by the correctness of the subject's plan to reach a destination in the neighborhood. The subject earns two points for making no error in the direction and the side of street of the destination, earns 1 point for getting one wrong and earns no point for getting both wrong. The subject will not walk into the traffic in any of the task evaluations in real streets.

COMSs who conduct real street evaluations will use paper data-entry forms to record objective performance assessment and subjective observation of subject performance. The original data-entry forms will be placed in the subject's data folder and secured in the PI's file locker after the data is transcribed to corresponding electronic forms and the transcription accuracy is versified by the PI or the research coordinator. Only computer generated subject IDs will be used on paper and electronic data forms.

COMSs who conduct real street evaluations will be blinded for subject training assignment. The objective nature of the performance measures ensures high inter-rater agreement, as shown in the investigators' pilot study. Nevertheless, COMSs of the UAB and AIDB sites will test the agreement of their evaluation outcome by evaluating the performance of the same subjects on real streets before the trial.

COMSs who are not involved in evaluation will train the active control group in real streets. The COMS will use their routine training procedure to conduct real street training to reflect the current practice in O\&M rehabilitation. COMSs have the freedom to customize their training procedure according to each subject's ability and need. However, the COMSs on UAB and AIDB who will conduct training will come to consensus on the exiting performance levels for the O\&M tasks and will confirm consistency of their exiting performance levels on the same low vision subjects before trial. The experimental group will be trained by the automated VR-IOMSs. A member of the research team will be in the room of the simulator for safety reason, but will not intervene VR-IOMS training. Members of the experimental group will also be asked to complete a survey of their learning experience with the VR-IOMSs. Members of the placebo group will watch low vision education video and discuss low vision issues not related to O\&M with COMSs.

COMSs who train subjects in real streets will keep an electronic training diary for each subject, which will include the participant's assessment of the subject's ability to perform the O\&M task entering and exiting training, observation of the subject's performance and progress during training. The VR-IOMSs automatically maintain detailed records of the teaching/practicing material, results of assessment, the errors made, the feedback/guidance provided and results of exit assessment. Subject will only be referred to using their study IDs in all training records.

Sample size computation for Task NLPT is based on the results of the investigators' pilot study, in which LV subjects received the NLPT surge training from COMSs. The mean pre- and post-training safety scores were 0.24±0.32 and 0.79±0.18. If the trial is designed to detect a 25% difference between the post-NLPT training safety scores of COMS and VR-IOMS training, and the enrollment is selectively stratified such that each group has an equal number of central and peripheral vision loss subjects, a sample size of 8 per group can achieve a power of 91% at a significance level of 0.05 in a one-way ANOVA study. Assuming that 20% of the enrolled subjects may not be able to complete the multi-session study, 30 LV subjects will be enrolled for the Task NLPT training trial. There have been no quantitative studies on TGJ training effect on LV patients. A quantitative study of 10 experienced LV travelers with mixed central and peripheral vision losses found a close match between their gap width judgments (safe gap width 4.6±0.54 sec) and cross times (0.48±0.88 sec). The authors' power analysis showed that the sample size, N=10, had 86% power to resolve a 0.5 second difference in the safety margin of the TGJ task. This resolution is good from a practical point of view (time for moving one step). the investigators will enroll 36 LV subjects (10 per group + 20% extra) for the TGJ VR-IOMS trial. No quantitative studies on LV patients learning the outdoor numbering system have been conducted. The investigators will use the subjects recruited for Task TGJ to conduct a pilot study to obtain training effects and standard deviations for sample size computation. For programmatic purpose, the investigators assume a sample size of 36 for the ONS VR-IOMS trial. This yields a total number of 102 LV subjects for the study.