Abstract
Current depth perception tests are either too time consuming or too coarse in resolution to detect subtle changes of stereoacuity. FInD-Depth is a rapid, self-administrable, and accurate method to measure stereoacuity across a range of stimulus conditions. We investigated FInD-Depth using noise and ring stimuli and compared stereoacuity estimates with 2-Alternative-Forced-Choice(2AFC) and clinical measures(Titmus and Randot) regarding test-retest reliability, thresholds, and test durations. FInD and 2AFC Depth used band-pass filtered ring and noise stimuli(1,2,4,8 cyl/š©). In FInD Depth, the stereoscopic disparity of a random subset of cells in three 4*4 grids were adaptively log-scaled from easy-(dā=4.5) to difficult-to-see(dā=0.1). Observers were given unlimited time to click on cells containing a target. In the temporal 2AFC task, observers indicated which of two 500ms intervals contained a target and a 3-down-1-up algorithm adjusted the stereoscopic disparity. 10 normally-sighted and 3 binocularly impaired observers were tested. For normally-sighted observers, no effects of test type nor spatial frequency for thresholds were found(2-Way-ANOVA). Mean stereo thresholds averaged across spatial-frequencies were 6.8, 7.2, 5.8arcmin for FInD Ring, FInD Noise and 2AFC respectively. Test duration was significantly longer for 2AFC(285s) than FInD(190s), Titmus(100s) or RanDot(93.1s). Test-retest biases were not significant: 0.34, 6.93 and -0.75 for FInD Ring, FInD Noise and 2AFC, respectively. Titmus results were highly correlated(r2>0.8) with 8cyl/š© stimulus but not for other spatial frequencies for all 3 tests whereas the Randot was not correlated with any spatial frequency or test. Stereoacuity deficits were detected by all tests in binocularly impaired subjects. FInD Depth stereo thresholds were comparable to standard clinical and psychophysical measures and significantly faster than conventional 2AFC measures. The method is able to test across a range of spatial frequencies in around 3min.