This study focuses on age-related macular degeneration (AMD) because, left untreated, it is the leading cause of blindness in the adult US population over 50. The goals of this program are twofold. First, the focus is to develop new screening tools to detect individuals with the intermediate stage of AMD, that are otherwise asymptomatic for vision loss, at a stage where they can be referred to a clinician for treatment and follow up, and when vision more likely can be preserved. The second goal is to develop Automated Retinal Image Analysis (ARIA) tools that help characterize Geographic Atrophy (GA), for which novel treatments are being investigated. We are developing data-driven algorithms for AMD screening which leverage recent advances in machine learning and machine intelligence and do not rely on detecting and counting or sizing drusen (a procedure which can be error prone). Instead, our method analyzes a fundus image as a whole using an image classification paradigm, and may also exploit ancillary patient information. Our goal is to use the AREDS dbGaP that includes thousands of images from hundreds of patients and offers a unique opportunity to develop and test these algorithms on a scale that has not previously been possible. Our team consists of the Johns Hopkins Wilmer Eye Institute, which will serve as the clinical analysis test site, and the Johns Hopkins University Applied Physics Laboratory, which will conduct the automated screening software tool analysis, development and testing.
The goals of this project are twofold. First, in order to work towards improving outcome, our goal is to develop novel software screening tools that allow for the automated detection of individuals with the intermediate stage AMD so that they may be referred to clinicians for follow up and treatment at an earlier stage than would otherwise occur. Second, we will develop new tools to characterize the evolution of GA, for which new treatments are being developed and tested. Our project is a secondary study on the AREDS dbGaP, which will be leveraged for the development and validation of these new algorithms on a scale never before attempted.
| Burlina, Philippe; Pacheco, Katia D; Joshi, Neil et al. (2017) Comparing humans and deep learning performance for grading AMD: A study in using universal deep features and transfer learning for automated AMD analysis. Comput Biol Med 82:80-86 |
| Feeny, Albert K; Tadarati, Mongkol; Freund, David E et al. (2015) Automated segmentation of geographic atrophy of the retinal epithelium via random forests in AREDS color fundus images. Comput Biol Med 65:124-36 |
