Abstract
Purpose :
Human color vision is achieved by mixing neural signals from cone photoreceptors sensitive to long (L-cones), medium (M-cones) and short (S-cones) wavelength light. In vivo mapping of the trichromatic cone mosaic provides the most direct and quantitative means to assess the role photoreceptors play in color vision and color blindness, but current methods are limited. Here we show using adaptive optics optical coherence tomography (AO-OCT) that photostimulation-induced optical phase changes occurring in cone cells allow their spectral types to be identified with unprecedented accuracy and speed.
Methods :
We imaged two color-normal subjects and one deuteranope subject with the Indiana AO-OCT system, collecting 5-s videos covering a 1°×1° patch of cones at 4° temporal to fovea. Halfway through each video, a 5-ms flash of visible stimulus (centered at 450 nm, 527 nm, or 638 nm) was delivered to the cones. Video acquisition with light stimulation was repeated 15 times with videos spaced 90 s apart, a total data acquisition time of 30 min. We extracted the phase difference between inner segment/outer segment and cone outer segment tip reflections of each cone cell in each volume. We classified the resulting phase traces into clusters using a k-mean clustering algorithm and labeled the clusters according to the expected spectral sensitivity relation for that stimulus: L>M>S at 637 nm, M>L>S at 528 nm and S>M>L at 450 nm.
Results :
Phase traces of cones in the two color-normal subjects exhibited a trimodal response distribution, regardless of stimulus. Comparing derived cone classes from the different stimulus conditions, the classification agreements between 637/528 nm, 528/450 nm and 637/450 nm were 97%, 94% and 94% for subject #1, and 97%, 93% and 93% for subject #2, respectively. Principle component analysis of the 637 nm results and Gaussian fits of the clusters (as defined in the AO photodensitometry literature) show our method can classify cones with an uncertainty less than 0.02%, at least two-orders of magnitude improvement over previous methods. Repeatability error was 0.3%. As expected, the phase response of cones in the deuteranope differed from the color normals, exhibiting a bimodal distribution with signal strengths expected of normal L and S cones.
Conclusions :
We have developed a highly sensitive and fast method for classifying cone types in the living human eye.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.