Browse Prior Art Database

A System and Method for the Early Prediction of Drusens Using the Combination of Retinal and OCT Images

IP.com Disclosure Number: IPCOM000248750D
Publication Date: 2017-Jan-05
Document File: 5 page(s) / 134K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is an automatic system that combines information from Optical Coherence Tomography (OCT) images with Color Fundus Photograph (CFP) to accurately estimate parameters of early stage drusen. The method combines the complementary features of OCT and CFP for accurate detection of early stage drusen for early prediction of age-related macular degeneration (AMD).

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 30% of the total text.

1

A System and Method for the Early Prediction of Drusens Using the Combination of Retinal and OCT Images

To detect early stage drusen for early prediction of age-related macular degeneration (AMD), Optical Coherence Tomography (OCT) gives height of the drusen; however, for small drusen the heights may be mistaken with noise. The Color Fundus Photograph (CFP) gives width and color of the drusen with clear pigmentation. Manual inspection of digital color fundus photographs (CFP) is the current gold standard for obtaining drusen parameters for evaluating the severity and monitoring the progression of AMD. However, inter-patient variation in fundus pigmentation, media opacity, retinal pigment epithelium (RPE), and choroid make the reliable localization of drusen on CFP challenging. Often, multiple readings from several graders are necessary in making meaningful assessments. While several methods can automatically identify drusen in CFP, the clinical usage of those methods is limited due to the above-mentioned causes. Screening systems that use those automated methods have to resolve to a coarse grading or need additional manual supervision (Gregori et al. 2014).

Spectral domain optical coherence tomography (SD-OCT) produces in vivo, high resolution cross sectional images of the retina. However, the RPE segmentation algorithms that come with OCT devices are not adequate for reliably picturing the anatomically correct amount of drusen and their total area (Schlanitz et al. 2010). While OCT imaging has recently emerged as a probable solution for diagnosis and progress monitoring of AMD, OCT imaging alone is not capable of accurately identifying all the drusen parameters (Yehoshua et al. 2013), (Jain et al. 2010).

The novel contribution is an automatic system that combines information from OCT images with a CFP to accurately estimate parameters of early stage drusen. The system takes as input a digital color fundus image and an OCT image of the same patient acquired at the same point/time, and then produces an accurate drusen map with parameters that correctly identify the small, early stage drusen parameters. The new method combines the RPE elevation map derived using the OCT image with the CFP image to predict an accurate drusen map. The drusen probability map can be generated in either two-or three-dimension (2D or 3D). In 2D, the drusen probability map highlights the height of the drusen (which is missing on fundus), which can be used for visualization and early diagnosis. The 3D drusen probability map provides the volume map of drusen and is useful for tracking the progression of the AMD.

To implement the system in a preferred embodiment: 1. On the request of the user, the system reads the OCT and CFP image pair from

the storage module 2. The following three flows simultaneously occur:

A. Flow 1 i. Compute a 3D surface of the RPE layer in the OCT image using a

2

learned Convolutional Neural Network (CNN) model ii. Use the RPE surface to calcula...