Abstract:

PURPOSE: The contrast sensitivity function (CSF) provides a comprehensive assessment of spatial vision in both normal and clinical populations. The shape of the CSF provides important information about the modulation transfer function and internal noise of the visual system (Chen et al., 2014; Hou, Lu, & Huang, 2014). Several visual diseases have been shown to differentially affect CSFs under different light conditions. In this study, we tested two hypotheses: (1) the shape of the CSF is invariant for each individual observer across different light conditions, and (2) the shape of the CSF is invariant across all the individuals.
METHOD: CSFs of 112 college students with normal or corrected-to-normal vision were measured using the quick CSF procedure (Lesmes, et al, 2010) in three light conditions (2.65, 20.2 and 95.4 cd/m2). The detailed experimental procedure is described in Hou et al, 2016. CSF is modeled by a truncated log parabola with four parameters: peak gain, peak frequency, bandwidth, and truncation level (Watson & Ahumada, 2005).
RESULT: To test hypothesis 1, we fit a model lattice to each observer’s data using a maximum likelihood procedure. We found that the model holding CSF shape parameters (bandwidth, truncation level) invariant across light conditions, but freely varied peak gain and peak spatial frequency across conditions, provided statistically equivalent fits as the most saturated model for 89.3% of the observers.To test hypothesis 2, we fit the aggregate data with a CSF model that held CSF shape invariant across all observers. This model, with 2 × 3× 112 + 2 parameters, is statically inferior to a model that allowed the shape parameters to vary across observers (with (2 × 3 + 2) × 112 parameters).
CONCLUSION: The results suggest that the shape of the CSF is invariant under different light conditions, but not invariant across observers.

Key words: Spatial Vision, Contrast sensitivity, Model comparison, Bayesian adaptive test, quick CSF