Abstract
Purpose :
To compare visual acuity measured on ETDRS logMAR using the descending limits method with a random presentation of line size and letter order (random logMAR). To compare these tests with a predictive threshold algorithm with resolutions of 0.1, 0.05 and 0.02 log units per line.
Methods :
Software was written in MATLAB R2014a. The ETDRS logMAR chart was reproduced and code written to randomize letter and line presentation order (random logMAR). These were compared with a new algorithm based on a double staircase method. The staircases are linked using a threshold prediction based on each prior response. 5 letters are presented from a set of Sloan letter permutations with standardized recognisability. The response is scored correct if 3 or more letters were read and incorrect if 2 or fewer were read correctly. The screen was calibrated using a Vernier scale and the maximum resolution was 0.02 log units. 8 thresholds were measured giving a mean and standard deviation. Three sets of tests were performed by two observers for induced refractive errors 0, -1 and -2 dioptres. Results of all three tests were compared.
Results :
With resolutions of 0.1, 0.05 and 0.02 log units and for refractive errors of 0, -1 and -2D the predictive algorithm gave mean acuities of -0.17, -0.16, -0.12, 0.07, 0.18, 0.16, 0.57, 0.49 and 0.48 log units respectively. Reproducibility was 0.05 for resolutions of 0.1 and 0.05 log units and fell to 0.15 for resolution of 0.02 log units. LogMAR and random logMAR acuities gave values of -0.18, 0.09, 0.41 and -0.19, -0.06, 0.28 log units. The differences between the tests had a mean of 0.04 log units but the predictive algorithm at 0.02 log units per line showed a difference of -0.16 log units in comparison with the logMAR tests. The standard deviation of acuity in the predictive test had a mean of 0.04 log units. The mean time for the logMAR tests was 52s and for predictive test was 110, 142 and 183s for 0.1, 0.05 and 0.02 resolutions.
Conclusions :
The predictive test provides 8 threshold measurements instead of a single measure of acuity and also a standard deviation. The reproducibility of the predictive test was excellent except at the high 0.02 resolution level. The predictive test allows multiple acuity measures in a single test whereas the logMAR and random logMAR give only a single measure and no error estimate.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.