Abstract
PURPOSE: To evaluate whether processing a color fundus photo (CFP) using an image embossing technique can improve the detection of reticular pseudodrusen (RPD). METHODS: This post-hoc analysis included the eyes of subjects enrolled in the Amish Eye Study with early or intermediate age-related macular degeneration and evidence of RPD. All patients underwent CFP, near-infrared reflectance (NIR), and fundus autofluorescence (FAF) imaging. The ground-truth presence of RPD was established with a combination of NIR and FAF imaging. An embossing processed (EP) image was created by replacing each pixel of the CFP image with a highlight or a shadow representing light and dark boundaries in the original CFP image. The presence of RPD in CFP and EP images was assessed by two graders in a masked fashion and the sensitivity of CFP and EP for detection of RPD was evaluated. Cohen's kappa (k) was used to test inter-grader agreement for CFP and EP. RESULTS: A total of 106 eyes from 62 patients with RPDs were analyzed. The sensitivity for detection of RPD on CFP and EP was 63.2% (95%CI: 52.0%-74.4%) and 91.5% (95%CI: 85.0%-98.0%), respectively. The inter-rater reliabilities of CFP and EP for RPD detection were 0.81 and 0.84, respectively. CONCLUSIONS: Embossing of CFP can improve the sensitivity for detection of RPD. The embossing technique can be a useful tool for better assessment of the true frequency of RPD in datasets where only CFP images are available.