Abstract
Purpose:
The only system used to map the human trichromatic mosaic in a living eye was an adaptive optics (AO) fundus camera at the University of Rochester. Since then, AO scanning laser ophthalmoscopes (AOSLOs) have been adopted widely, owing to their increased resolution and contrast. Our goal was to investigate the feasibility of employing an AOSLO for objective spectral classification of the human cone mosaic.
Methods:
An AOSLO was used to image the cone mosaic at 543nm wavelength. The retina at 4 deg eccentricity was dark-adapted for 5 min to maximize photopigment concentration. Five second AO-corrected retinal videos were obtained after dark adaptation or after a selective bleach of L-cones at 680nm following dark adaptation. The intensity timecourse of each cone was measured from 5 stabilized videos and fitted with an exponential function to yield individual cone bleaching curves. First, the difference of the fully bleached and dark-adapted intensities was plotted on a histogram and fit with sum of 2 Gaussians. The intersection of the Gaussians delineated likely S-cones, which were confirmed by inspection of their individual bleaching curves. Next, for each remaining cone, the intensity difference between its fully bleached and selectively bleached state was plotted against the intensity difference between its selectively bleached and dark-adapted state. The polar angle for each data point was calculated, plotted in a histogram and fit with sum of 2 Gaussians. The intersection of the Gaussians delineated L and M cones. The ratio of the sum Gaussian and the comprising Gaussians at any polar angle denoted the probability whether a given cone was L or M.
Results:
Following dark adaptation, imaging with 543nm led to depletion of photopigment and increase in mean image intensity by 50%, indicative of the available photopigment concentration. Out of total 303 contiguous cones identified, S-cones accounted for 5.9%. The percentage of misidentified L or M cones was 4.6%. Among the cones outside the misidentified overlap area of the Gaussians, the average probability of identification of L and M cones was 0.93. The resulting L:M cone ratio was 1.5:1.
Conclusions:
The objective spectral classification of the cone mosaic is demonstrated with an AOSLO. Classifying the cone mosaic in this manner puts us in an ideal position to harness the advantages of the AOSLO in unraveling the circuitry between visual perception and individual photoreceptors.
Keywords: 471 color vision •
551 imaging/image analysis: non-clinical •
648 photoreceptors