Abstract
Purpose: :
To study subtle aging changes in the retinas of normal subjects, and to determine which features are enhanced by using polarization analysis.
Methods: :
Forty subjects (20 aged 20 - 29 yr, 20 subjects 70 - 79 yr) were imaged with macula- and ONH-centered fields of view, using a modified scanning laser polarimeter (GDx) with 780 nw illumination and 15 x 15 deg field of view. Five types of images were computed from each dataset (Matlab) that varied polarization content. Each dataset contained a series of images that varied in input polarization; light returning from the eye was divided between parallel and crossed detectors. (1) The average of all conditions and for both detectors produced a confocal image. (2) The depolarized light image was computed from the minimum gray scale value for each pixel at the crossed detector. (3) A scattered light ratio image provided normalization of the scattered light by dividing by the average light return, thereby minimizing the effects of refractive index changes or uneven illumination. (4) The modulation of the crossed detector at each pixel over polarization condition produced a birefringence image. (5) A new image type, the birefringence ratio image, was computed by dividing the birefringence image by the average light return, again to remove the variation due to refractive index changes or uneven illumination. Grading was blind to age, and rated scattered light, RPE changes, arteriole and venule changes, peripapillary atrophy, peripapillary hyperpigmentation, myopic crescent, and NFL loss.
Results: :
The ratings differed significantly among features graded (P < .0001), with scattered light > RPE changes > vessel changes, but individual differences were usually larger than the differences between young and old for most image types. Peripapillary atrophy was rated as more visible in the depolarized light image and slightly less in the scattered light ratio image, compared with other image types. Confocal images of maculas differed for younger vs older subjects (p = .032).
Conclusions: :
Polarization content enhances certain features more than others. Individual differences among normal mask aging changes. Specific image types prove promising for detecting actual pathology.
Keywords: image processing • aging