Neural limits of visual resolution

Abstract

Aims: The aim of this thesis was to formulate a model of retinal neural circuitry in humans based on morphological estimates of retinal cone and ganglion cell populations, to compare results based on this model to psychophysical measurements of visual resolution and identify the retinal neural factors that limit visual resolution.Materials & Methods: Total displacement of central ganglion cells from foveal cones was obtained by estimating the length of fibers of Henle in four human retinas along the nasal or vertical hemi-meridians. Quantitative estimates of ganglion cell and cone densities were made using a modified disector method in vertically sectioned human retinas and ganglion cell densities were adjusted for central displacement. High-pass Resolution Perimetry (HRP) thresholds were used as a psychophysical estimate of visual resolution throughout this work. Functional magnetic resonance imaging (fMRI) estimates of the human linear cortical magnification factor were obtained from the literature. Resolution and detection thresholds of computer generated high-pass filtered optotypes were recorded under equal lighting and luminance conditions.Results: We obtained quantitative estimates describing total displacement of ganglion cells from cones within the central 3 mm (11 deg) of the human retina and established a model of the retinal vertical pathway from cones to ganglion cells based on quantitative estimates of cone and ganglion cell densities as well as calculated effective ganglion cell densities. Center-to-center separation of the effective sampling area of one midget ganglion cell independently of its type (ON or OFF), i.e. the estimated receptive field, proved to be a good predictor of visual resolution within the central 20 deg of the human retina. Strong linear relationships through the origin were established between ganglion cell separations and HRP thresholds. The overrepresentation of the fovea and immediately surrounding retina in the human striate cortex was confirmed to be the result of an increase in devoted cortical distance per central ganglion cell. A dependence on contrast was found with and without correction for peripheral refractive errors for both resolution and detection thresholds of three high-pass filtered optotypes, irrespective of the task being resolution or detection. A linear proportionality through the origin was shown between resolution and detection thresholds and ganglion cell separation.Conclusions: Presented results of lateral displacement support and extend previous analyses of the relation within the human fovea between quantitative cell distributions and psychophysical measurements. The separation of estimated receptive fields can be interpreted as a limiting factor of visual resolution in the human retina. Intra-individual comparisons of HRP thresholds and ganglion cell separations support the hypothesis of a direct proportionality through the origin between retinal ganglion cell separation and HRP thresholds. There is a non-linear scaling between central retinal subunits and dedicated striate cortical distance in man, with an overrepresentation of the fovea and immediately surrounding retina in the striate cortex. None of the investigated high-pass filtered optotypes were found to be contrast limited irrespective of the task being resolution or detection. This result is quite the contrary of that reported for gratings and indicates that classical sampling theory may need modification before it can be applied to high-pass filtered optotypes.