An Evaluation of the Agreement Between a Computerized Stereoscopic Game Test and the TNO Stereoacuity Test

Abstract

Purpose: Stereo-anomaly is commonly associated with amblyopia. An investigation was conducted to determine whether the measurements of stereoacuity obtained with the stereoacuity reference test (TNO Test) show an agreement with a computer stereoscope video game.

Methods: Thirty-two subjects (mean age 9.37±2.00 years) with an amblyopia history were selected for a blind and randomized study of stereoacuity improvement through a new random dot game. A masked examiner measured the stereoacuity three times per subject using the TNO test (at the beginning, at the end and after 6 months of the treatment). A second masked examiner measured stereoacuity using the new computerized game after the TNO masked evaluation.

Results: The Pearson's correlation coefficient one test against the other was r² = 0.767 and the Bland-Altman plot was r²= 0.069 (mean difference -0.03 log sec). Using three categories: poor (840-300 seconds of arc), coarse (480-210 seconds of arc) and moderate-fine stereoacuity (210-30 seconds of arc). Positive predictive values were 89.5% for moderate-fine; 72.7% for coarse; and 90.0% for poor stereoacuity. In addition, the agreement was evaluated using the Kappa coefficient (K= 0.743) with a 0.95 confidence interval and lower and upper Kappa limits were (0.628 and 0.858), respectively. Kappa coefficient and limits were still good when analyzing data before (K =0.663, 0.420 and 0.906) and after the treatment (K= 0.765, 0.632 and 0.899).

Conclusion: The Computerized Stereoscopic Game test allows the measure of stereoacuity. It can be used for both the purpose of detecting stereo vision deficits or tracking stereo vision development.

Keywords: TNO test; amblyopia; computerized game test; gamification; stereoacuity.

Figure 1

Logical process of the game. The subject must identify the hidden silhouette, selecting one from those shown in the bottom part of the screen (left image). If the answer is correct, the software plays a high-pitched sound and the same image appears in the form of picture (right image). If the subject provides three consecutive correct answers, the software generates a new screen with a random dot image representing a finer stereopsis. If the subject provides a wrong answer, the software plays a deep sound and the random dot image does not change (left image). Finally, if the subject provides three consecutive wrong answers, the software shows the right answer (right image).

Figure 2

(A) Pearson’s correlation coefficient obtained through the comparison of measurements obtained by the CSG vs TNO test, both in logarithmic values. (B) Bland–Altman plot that compares stereoacuity logarithmic values measured by the CSG vs TNO test. The dashed lines show the limits of agreement for a 95% prediction of the population and the dotted line the mean difference between the methods compared and the confidence interval of the mean.