Abstract
Abstract: :
Purpose: To investigate light–tissue interaction in normal eyes and those with drusen, using scanning laser polarimetry. To investigate methods to automatically discriminate between normal subjects and patients with drusen due to age related macular degeneration. Methods: Polarimetry images were obtained in twenty normals ( age:17–55y, 10 female, 10 male) and 20 drusen patients (age: 53–86; 11 female, 9 male) using a scanning laser polarimeter (GDx). Macular centered images (15x15 deg field, 780nm light) were acquired with a series of 20 input polarization angles. Returning light was collected by two detectors separated by a polarizing beam splitter. We used customized software written in Matlab (Mathworks) to compute three different images: a depolarized image, computed as the minimum of the crossed detector; an average image, which is the average of both detectors; and a ratio image, which are generated from the ratio of the depolarized to the average image. We hypothesize that retinal pathology that leads to increased scattered light would lead to a higher mean value for these ratio images. Using Matlab we calculated the mean intensity, SD, and CV for a 213x213 pixel field excluding the center 30x30 pixels, which removes an instrument artifact. We filtered the ratio images using a 3x3 or 6x6 gaussian kernel. Results: In 19 of the 20 images of the normal subjects, a higher intensity was found in the center region extending up to 10 deg. This was also seen in some of the patients with drusen. In the ratio image, drusen appear dark with bright rims. Drusen are readily visualized in the depolarized image, but due to noise reduction can be of more striking contrast in the ratio image, which might deemphasize small pathological features. Quantitatively, mean brightness was lower in normals than in drusen patients (0.134+0.34; 0.294+0.091; p<0.0001), whereas the CV was higher in the normals (p< 0.0001). All but one drusen patient had higher mean intensity in the ratio image than would be predicted by a linear function fit through the age vs. intensity data of young to middle–aged subjects. Conclusions: Three image types emphasize differences between normal subjects and patients with drusen, but in different ways. Gaussian filtered ratio images did not help to remove sufficient noise to permit automatic discrimination based on simple summary statistics. The central region of relatively higher intensity in depolarized light images is a dominant feature in healthy eyes and influences the quantitative measurements.
Keywords: age-related macular degeneration • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)