Evaluation of a nonlinear diffusion process for segmentation and quantification of lesions in optical coherence tomography images

Abstract

We evaluate the ability of a nonlinear anisotropic diffusion process to enhance the contrast for structural boundary regions and to reduce the speckle noise in optical coherence tomography (OCT) images. We also investigate the suitability of various image features, such as gradient magnitude and intensity or gradient profiles, for boundary localization. The results suggest that the nonlinear anisotropic diffusion method has potential in assisting segmentation and quantification of fluid-filled regions in clinical OCT images.