Abstract
Purpose :
Vision tests within a screening battery should be quick, easy and independent. Because the early visual system has limited functional architecture, many vision screening tests ultimately provide redundant information. For example, while seeking information about parallel processes such as stereopsis and motion perception, certain test battery results may be muddied by colinear relationships and shared computations. These redundancies potentially make visual screening batteries inefficient and less informative. For career fields requiring extensive medical screening (e.g. commercial pilots/drivers, military aviation, etc.), the duration of the screening may be an important factor. However, by identifying independent factors underlying visual performance, a test battery can be streamlined. The goal of this dimension reduction analysis is to identify such underlying factors in a visual screening paradigm from a large and diverse subject population.
Methods :
192 subjects underwent the Operational Based Vision Assessment Laboratory Automated Vision Testing (AVT) procedure, which included computer-based tests for visual acuity, luminance and cone contrast sensitivity, motion coherence, stereopsis, and binocular oculomotor function. Psychometric thresholds and fusional vergence ranges were collected from each subject. Factor analysis was performed on a total of 14 normalized variables with a promax rotation.
Results :
This rotation identified 5 factors that explained 74% of the variance in the dataset. These factors were related to the following: 1) medium and high spatial frequency vision, 2) stereoacuity and horizontal fusional range, 3) color vision, 4) motion perception, and 5) low spatial frequency vision. Variability within individual factor loadings revealed informative trends. For example, while horizontal fusion range was associated with factor 2, vertical fusion range was not well correlated with any factor.
Conclusions :
These results suggest that the number of tests within the AVT battery may be reduced to as few as five with a limited loss in information. Furthermore, identifying these underlying factors may have additional utility in predicting performance in naturalistic or operational visual tasks.
This is a 2021 ARVO Annual Meeting abstract.