Abstract
Purpose: :
To evaluate foveal localization variability in patients with non-exudative age related macular degeneration (AMD) patients and age-matched normals using custom scanning laser polarimetry images. To determine if increasing levels of pathology negatively impact foveal localization.
Methods: :
A scanning laser polarimeter (GDx, LDT/CZM) was used to acquire 15 degree macular images (256 x 256 pixels) in 25 patients diagnosed with non-exudative AMD without central geographic atrophy (ave age 69.8, sd 6.5) and 25 age-matched normals (ave age 70.2, sd 7.0) . Raw image data were used to compute 5 custom images differing in polarization content: confocal image, depolarized image (emphasizing deep pathology), maximum parallel detector image (emphasizing highly reflective structures), 2 birefringence images (1 with birefringence magnitude and orientation information and 1 with magnitude information only). The fovea location was determined by 3 graders. Paired t-test was used to compare the variability in foveal location between AMD patients and normals. Variability was also compared across image types. A 5 x 5deg region of interest centered on the fovea was used to examine central pathology in the depolarized image (areas at least 3 x 3 contiguous pixels above or below 2 st dev of the depolarized image mean). The amount of pathology was compared to variability in foveal localization using linear regression.
Results: :
AThe variability of foveal localization was greater in AMD than in normals. (paired t-test, p<0.01). The birefringence image with magnitude and orientation information was least variable for both AMD and normal subjects (ANOVA, p<0.01 for both). The overall average inter-grader distance in the birefringence image with orientation was less than 8 pixels (141 microns) for AMD subjects. Planned comparisons showed no statistically greater variability between the 2 birefringence images, but the birefringence image with orientation was statistically less variable than the additional 3 image types graded. The area covered by pathology in the depolarized image was not correlated with foveal localization variability in the birefringence image with orientation (p=0.12, r2=0.102).
Conclusions: :
The fovea can be located with low variability in images containing birefringence information in normals and patients with AMD. The fovea can be accurately localized even in the presence of significant retinal damage, which often precludes accurate localization using other imaging techniques.
Keywords: imaging/image analysis: clinical • age-related macular degeneration • retina