Abstract
Purpose: :
To determine whether the fovea can be located in near infrared images, which lack typical landmarks such as strong macular pigment absorption. These images are also too small to include the optic nerve head or the entire vessel arcades as landmarks. To determine the agreement among observers of foveal location measurements using images obtained from a confocal scanning laser polarimeter.
Methods: :
We used confocal scanning laser polarimetry (GDx, LDT/CZM) to digitize a series of 780 nm macular images of 256 x 256 pixels and 15 deg. There were 40 raw images: 20 input polarizations and 2 detectors, one parallel to the input polarization and one perpendicular (crossed). From the raw images we computed 18 images with differing polarization content from little to most: a depolarized (scattered) light image from the crossed detector, an image from the average of both detectors, two image types from the parallel detector, and two birefringence image formats that emphasize the macular cross. Three graders independently recorded the foveal location in 6 of the 18 image types obtained from 35 normal subjects, age 19–90yr. The foveal location was based on retinal vessels, retinal pigmentation changes, the foveal reflex (when visible), and the macular cross, all of which could be compared to an adjustable ring of 4.5 deg diameter. The coordinates of the foveal location were averaged for the three graders. The standard deviation from the mean was calculated to determine inter–investigator agreement.
Results: :
The agreement among observers was surprisingly good, < 2 pixels, 25 microns or less, for all except the depolarized light image, which was 3.5 pixels on average. The depolarized light image had significantly worse agreement than all of the other image types (p<0.005), with the agreement for the other image types not statistically different (p>0.28 or more). The images for different subjects did vary in the ease with which the graders were able to locate the fovea, but typically the agreement indicated that the fovea was located with sufficient accuracy for clinical purposes.
Conclusions: :
All three of the graders were able to locate the fovea in the 15 deg near infrared images, despite the lack of typical retinal landmarks.
Keywords: imaging/image analysis: clinical • macula/fovea • clinical research methodology