Abstract
Abstract: :
Purpose: To automatically detect wedge shaped defects in polarimetric nerve fiber layer (NFL) images, which might improve automated glaucoma diagnosis. Methods: The NFL images were acquired with a prototype GDx (Laser Diagnostic Technologies, Inc., San Diego, CA), equipped with a variable corneal compensator. We developed an algorithm that transforms the polarimetric image into a polar representation, searches for edges extending to the optic disc and combines opposing edges to form the detected wedge. The edge locating scheme employs a modified dynamic programming technique and favors straight edges. We used 8 images with visible wedges and 37 images without wedges as a test set. The result of the algorithm was compared with expert clinical judgement.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, S • imaging/image analysis: non-clinical • retina