Abstract
Purpose :
To derive the empirical frequency-of-seeing (FOS) curves in patients with glaucoma from longitudinal visual field data using a probabilistic modeling.
Methods :
Series of 15 visual fields were taken from 130 glaucoma patients in the Rotterdam Eye Study (in the public domain). Smooth regression was used to prepare test-retest data for a hypothetical glaucoma patient. The author modeled the test-retest data as a multinomial distribution generated from a biased die with 42-faces (<0, 0, 1, ..., 40dB). FOS curves, defined as cumulative gaussian function, were incorporated into the bias, based on the possible response sequences at four primary points (±9°, ±9°) during the Humphrey Full Threshold testing. The parameters of FOS curves (luminance of inflection point, slope, and asymptotic maximum response probability) were estimated by a computer program written in the probabilistic programming language stan.
Results :
FOS curves corresponding to each sensitivity from 0 to 35dB were successfully derived. As glaucoma progresses, the asymptotic maximum response probability declined monotonically from 99% (35dB) to 14% (0dB). In contrast, the horizontal shift of inflection point stopped spontaneously when the luminance reached approximately 19dB. Consequently, the psychophysical threshold for luminance increment was not changed substantially in moderate to severe stages (Fig. 1).
Conclusions :
These results show that visual dysfunction in glaucoma patients is distinctly different from previously proposed. Interestingly, the luminance increment threshold appears to be inherently limited.
This is a 2021 ARVO Annual Meeting abstract.