![[Previous Chapter]](/icons/button-back.gif){kind=icon}
![[Table of Contents]](/icons/button-contents.gif){kind=icon}
![[Next Chapter]](/icons/button-next.gif){kind=icon}
| Publications Page | HITL Home |
The purpose of this experiment was to evaluate a fundamental concept underlying a potential rehabilitation technique for use with patients suffering from chronically low VOR gain. The concept, suggested by Viirre (1996), attempts to �coax� VOR gain upwards using several small incremental increases in gain demand (e.g., 5 to 15%) instead of one large increase (50 to 100%). VOR gain mechanisms can work quickly and efficiently to compensate for small gain-change demands but are not able to ever completely compensate for large demands (Collewijn, et al., 1983). Patients with chronically low VOR gains face the latter problem in everyday life; there is too great a difference between their current gain level and normal levels which continually present a single, large, unachievable demand. If virtual interfaces are able to modulate VOR gain levels, perhaps this one large demand could be replaced by several smaller, more manageable demands experienced virtually.
The previous three experiments demonstrated that virtual interfaces can indeed modulate VOR gain levels and the first experiment demonstrated a link between image scale and change in VOR gain direction/magnitude. Therefore, small increases in VOR gain demand could be provided by slightly adjusting GFOV from the neutral condition (where GFOV = DFOV).
This experiment investigated an underlying premise supporting this proposed rehabilitation technique; i.e., that a series of small gain-change demands could result in higher overall VOR gain adaptation than one large gain-change demand over the same period of time. This experiment addressed main objective 4 of this dissertation (See Chapter 1).
There are many issues to consider even with this one underlying premise. First, there is the question of how many increments to use and how big a gain change demand to provide with each increment. Should these demands be fixed in magnitude across increments or variable? Also, what are the criteria for switching from one step to the next? Should it be after a set period of time or should the subject�s level of compensation dictate matters? These are all valid questions for which there are no solid answers. As a result, the configuration of this experiment represented initial conjectures. Subjects were provided gain change demands that varied slightly in magnitude over a series of 5 fixed intervals during a 30 minute exposure.
This experiment explored VOR gain increases instead of gain decreases for two reasons. First, gain increases are in the appropriate direction to aid those with chronically low VOR gains. Second, it was anticipated if the incremental method was effective, it would be more readily noticed in this condition (due to the lower magnitude gain changes observed in this direction).
It was hypothesized that the incremental step method would result in a larger VOR gain increase than a single, large gain-change demand. The hypothesis is unavoidably qualitative at this point given the exploratory nature of this experiment.
Five subjects (3 male and 2 female, age range: 23 to 39) volunteered to participate in this experiment. All reported to be in good health with no history of visual-vestibular medical problems. Four subjects also participated in the Image Scale Experiment, so although they were not new to the virtual interface used, they had not experienced it in at least the previous 10 days. The fifth subject had not previously been tested.
A two factor within-subjects design was employed, with two levels of DEMAND (single, step) and two levels of TESTTIME (pre, post). Given that four subjects previously participated in the Image Scale Experiment and had already received the equivalent of a single gain increase demand stimulus (the MAG condition), they only needed to experience the incremental method of gain adaptation. DEMAND was thus confounded with session order, but this appears to be a minor issue given that past studies did not demonstrate any statistical effect of session on percent gain changes. The fifth subject was run only in the incremental method. Though the same three testing frequencies were used (0.2, 0.4, and 0.8 Hz), these were not separated out in the analyses (generalized adaptation had already been demonstrated in the three previous experiments).
The dependent variables were VOR gain and phase estimates, averaged across two trials per cell as before.
8.4 Experimental Set-up and Apparatus
The set-up was as in the Image Scale Experiment (Section 5.4) except for the following changes. Subjects in the step condition received 5 virtual images as before (approximately 6 minutes per VE), but the image scale was increased for each successive VE presentation. Image scale began at 1.09X and progressed through 1.26X, 1.5X, 1.71X, and finally to 2.0X for the final image. This contrasts with the single demand condition, when subjects received an image scale of 2.0X for the entire 30 minutes. Image scale changes were driven by changing the GFOV in the WARP TV software. The minimum system time delay (48 ms) was employed throughout.
The procedure was as in the Image Scale Experiment (see Section 5.5) except for the following changes. No re-adaptation testing was conducted and no sickness testing was recorded (other than oral reports during, which were recorded as part of the safety protocol).
VOR data were analyzed as in the Image Scale Experiment.
VVOR data are shown in Table 23. The virtual world VVOR data were collected with a image scale of 1.09X. VVOR phase showed an increase in lag in the VR condition.
|
Viewing Condition |
VVOR gain |
VVOR Phase (deg) |
|
Real World |
0.93 |
-3.1 |
|
VR (1.09X) |
0.98 |
-18.4 |
VOR gain and phase data by DEMAND are presented in Table 24. The subject who only participated in the STEP experiment did not show any gain change as a result of exposure. For ease of comparison, the STEP condition is shown both with and without his data. Only the STEP condition with "N = 4" was used for statistical comparison with the corresponding SINGLE condition.
A one-tailed paired t-test indicated a strong trend toward an increase in gain within the STEP condition (t(4) = 2.06; p < 0.06). After verifying homogeneity of variance (F = 0.24; p > 0.11), a two sample t-test was performed comparing the two conditions with no statistical differences found (t(3) = 1.94; p > 0.60). The entire data set from the MAG condition is also included in Table 24 for comparison purposes.
An informal comparison of DEMAND by subject found that two subjects had a greater gain increase in the STEP condition and the other two subjects had a higher gain increase in the SINGLE condition.
This experiment served as an initial investigation into the nature and value of the incremental step method. Due to the sample size used, only very large differences between the two methods would be capable of reaching statistical significance. However, the goal of this experiment was to search for trends favoring one technique over another and also to gather knowledge that may help optimize the STEP technique for adaptation.
The VVOR gain data indicated once again that varying GFOV can create a gain change demand on the VOR adaptive system. In this case, the commanded gain change for the VR VVOR was small (approximately a 9% increase) and the VVOR gain response was more nearly compensatory (6% increase). The phase lag became greater in the VR VVOR, which has been previously hypothesized to be due to the effects of the minimum time delay along with the difficulty of tracking a virtual target that is stationary in virtual space but moving relative to real space.
The version of the STEP method used in this experiment did cause a VOR gain increase to occur that nearly reached statistical significance (at the 0.05 alpha level) after testing only 5 subjects. This indicates that it is at least a viable alternative technique for increasing VOR gain.
The comparison between the STEP and SINGLE methods were inconclusive. The failure of the difference between the methods to reach statistical significance is of minor importance given the small sample size. However, there appeared to be no trends favoring one method over the other. Looking only at the four subjects that used both methods, more adaptation occurred in the SINGLE method (8% increase versus a 5% increase). Expanding the sample sizes to include all subjects who used either method, the results still favor the SINGLE method but the difference between the two methods is reduced (5.9% versus 4.6%).
However, this emerging trend falters upon more investigation. Of the four subjects who used both methods, half performed better with the STEP technique. Additionally, one of the four data points (a data point being the average overall percent gain change per subject) in the SINGLE condition was twice as large as any other data point in the STEP or SINGLE condition. This value was 23% gain increase and was not considered an outlier when compared to all the subjects who participated in the MAG condition. Therefore, it was not excluded from the analysis. However, this value disproportionally influenced the average gain increase in the SINGLE condition. Without it, the average gain increase in the SINGLE condition is 3.65, which is lower than the STEP average.
Lastly, the baseline gain of the VOR varied (within the same group of four subjects) between the two methods by close to 10 %. This resulted in a significantly higher pre-exposure gain in the STEP condition which could potentially have effected the results (e.g., if the subjects were overly alert during pre-gain testing in the STEP condition, larger increases in gain could be masked by the elevated pre-gain estimates). However, an individual�s baseline VOR gain estimates are known to be variable over time, so the most accurate procedure for assessing VOR adaptation is to utilize the baseline gain measured just before the exposure period begins (versus some averaged measure over a longer period of time). Larger sample sizes would likely reduce this baseline gain discrepancy.
The results of this study are therefore inconclusive as to the relative benefits of the two methods. However, these results have demonstrated that the STEP approach is a valid method for increasing the gain of the VOR. Furthermore, the specific configuration of the STEP method employed in this experiment is likewise supported. These results, combined with Viirre�s work with clinical patients (Viirre, Draper, Gailey, Miller, & Furness, In Press) provide a base of support from which to further study this method. Future experiments should concentrate first on optimizing the STEP method and then comparing the optimized version to the single method.
A few words on the phase data are in order. The phase appeared to decrease (lag) more substantially in the STEP method (as compared to either the SINGLE method: 4 subjects or the entire sample from the MAG condition). Assuming that these results are confirmed by further research, this presents an interesting question. Why would phase adaptation more readily occur in the STEP method when the only parameter varying between the two methods is visual scene amplitude? The only plausible explanation is that phase adaptation is somehow adversely influenced by increasing amplitude of visual scene motion. A discussion of this possibility is presented in the next chapter.