Abstract
Purpose :
To evaluate the feasibility of using signal detection theory (SDT) to examine the ocular surface sensory processing of corneal pneumatic mechanical stimuli using the Waterloo Belmonte aesthesiometer.
Methods :
Ten asymptomatic participants were recruited in this study. All participants were naïve to the SDT part of the experiment, but 6 participants were familiar with the other experiments using esthesiometer. The pneumatic stimulus was at eye temperature (approx. 33°C) and to change strength, flow was systematically varied. The stimulus intensities for the SDT experiment were obtained using the ascending method of limits, in which three ascending runs were averaged to estimate the threshold. Then, to keep the signal constant for each participant, flow was set at 1.5X threshold. 100 trials (demarcated using auditory cues) were presented with a signal probability of 0.4 (i.e., 60 catch trials with stimulus intensity of zero) and with signals and catch trials in random order. Responses (“yes” (there was a signal) or “no” (there was no signal)) were recorded using a button box with auditory feedback for each response. Breaks occurred after approx. 50 trials. Participants were instructed to blink normally between stimuli, and training was provided before the data were used. Detectability (d’), criterion (c) and likelihood ratio (β) were calculated using false alarms and hit rates.
Results :
The average (± SE) d’ was 0.52 (± 0.12), criterion was 0.22 (± 0.10) and β was 1.11 (± 0.06).The area under the curve obtained was 0.64 (±0.03). In addition, non-parametric SDT metrics were calculated. The average A’ was 0.67 (±0.03) and the non-parametric response bias β” was 0.07 (±0.03).
Conclusions :
The experiment shows the feasibility of using a SDT approach to examine ocular surface sensory processing with less concern about the potentially important impact of participant criterion changing threshold and masking meaningful effects. The data suggest that participants in this experiment chose a relatively conservative strategy (reporting ‘no’ to trials more commonly) but this might be anticipated considering we designed the experiment with a relatively large proportion of catch trials. The data also suggested that some of the assumptions about classical SDT (underlying homoscedastic Gaussian ‘signal + noise’ and ‘noise’ distributions) need to be examined.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.