Validation of Angular Indication Measurement (AIM) Stereoacuity

Abstract

Background: Stereopsis is a critical visual function, however clinical stereotests are time-consuming, coarse in resolution, suffer low test-retest repeatability, and poor agreement with other tests. We developed AIM Stereoacuity to address these limitations and asked whether it could deliver reliable, efficient, and precise stereo-thresholds across stimulus types and disparity signs.

Methods: Observers reported the orientation of 5×1.25° bar defined by disparity of random dots embedded in a 6° diameter circular cell, presented in a 4*4 grid. Bar disparity was scaled from ± 2σ relative to a threshold and slope-estimate, initially set by the experimenter and adaptively updated. Orientation report errors (indicated vs. actual bar-orientation) were fit with a cumulative Gaussian function to derive stereo-thresholds. Twenty-one normally-sighted observers were tested with red-blue anaglyphs in crossed and uncrossed disparity signs across 4 element-types (8.5arcmin broadband dots, or band-pass difference of Gaussians with peak Spatial-Frequency (SF) of 2, 4, or 6 c/°). We analyzed stereoacuities, test durations, and the test-retest repeatability.

Results: Across SFs and observers, test duration for a chart were 36 and 40 secs for measuring crossed and uncrossed disparity, respectively. There was no effect of disparity sign or SF (Kruskal-Wallis; p>0.05). Median log stereo-thresholds averaged across all SFs were 1.90 and 1.84 log arcsec for crossed and uncrossed disparities, respectively. Crossed and uncrossed disparities were moderately correlated across SFs(r=0.44 to 0.79; median=0.54). Test-retest biases were 0.01 arcsec (p=0.45) and 0.10 arcsec (p= 0.001) for crossed and uncrossed disparities, respectively.

Conclusions: The results for the response-adaptive, self-administered AIM Stereoacuity method showed no significant stereo-thresholds differences between broad- and narrow-band stimuli. The test delivers repeatable results for crossed disparity in approximately 80 seconds.

Keywords: Depth perception; binocular vision; response-adaptive psychophysics; stereoacuity.

Figure 1)

AIM Stereoacuity’s stimulus for dot elements with broadband spatial frequency. The range of disparities presented on the first chart was defined by the experimenter (1° to 1’ in even log-spaced steps) and on subsequent charts was adaptively calculated based on the threshold and slope of each participant’s responses to previous charts. The illustration in grey visualizes the varying degrees of depth due to the varying degrees of horizontal disparity. B) Orientation report error (y axis) as a function of the stereoscopic disparity of the bar (x axis) for a typical naïve participant. Orientation errors (blue circles) for each of the 48 indicated cells as a function of horizontal disparity were fit with a cumulative Gaussian function (equation 1, red curve, ±95% CI green dashed lines) to derive the stereo-threshold.

Figure 2:

A) Time required to complete a single Stereoacuity charts for crossed and uncrossed disparities. 2B-C) Stereoacuities measured with AIM Stereoacuity for broadband (0c/°) or band-pass (2, 4 or 6 c/°) elements. Datapoints show the results for individual participants averaged across the first and second test for crossed (B) and uncrossed (C) sign. Horizontal lines show the median, boxes show the interquartile range and whiskers show 95% intervals.

Figure 3:

Correlation between crossed (C) and uncrossed (U) stereoacuities (log arcsec) for broadband (0) and SF bandpass (2, 4 or 6 c/°) elements. The dots show data for individual observers, the line shows regression. The bar charts in the diagonal show the frequency distribution for each SF condition. The Kendall’s rank correlation coefficients and p values for each condition are shown in the circles.

Figure 4:

Bland-Altman plots to assess the repeatability between AIM Stereoacuity tests from 2 runs for crossed (A) and uncrossed (B) disparity stimuli. All participants test-retest difference in log stereo-thresholds are shown as circles. The black solid lines indicate the zero difference (as reference) and the upper and lower limits of agreement, the black dashed lines show the mean bias. Figure 4C & D: Stereo-threshold estimates following each chart in the first and second session for crossed and uncrossed disparities. Each line shows individual participants, the bold lines indicated the median for each chart number.

Figure 5:

A) Orientation report bias as a function of target orientation for crossed stimuli. The x axis shows the orientation of the target bar (0° = horizontal), the y axis shows a histogram of report errors. The bin size was 5°, the colorbar indicates the frequency of error occurrences. B) Fourier analysis of orientation error. C) Mean error and kappa for crossed stimuli. D-F show similar figures for uncrossed stimuli.