Utility of Deep Learning Methods for Referability Classification of Age-Related Macular Degeneration

Abstract

In an extension of previous work, 1,2 we assessed 2 deep learning (DL) methods addressing a 2-class age-related macular degeneration (AMD) referability classification: referable for the intermediate or advanced stage of AMD or not referable. Methods | We used 67 401 color fundus images (keeping only 1 image for each original stereo pair) from the National Eye Institute Age-Related Eye Disease Study (AREDS) data set 3 that were taken from 4613 individuals (who provided written consent) over a 12-year study, including baseline and follow-up visits, from November 13, 1992, to November 30, 2005. 1,2 The original AREDS image grading from certified graders at a fundus photograph reading center 3 were used as the gold standard. The present analysis was performed from January 22, 2018, through April 19, 2018. Use of the AREDS data set was performed following Johns Hopkins University School of Medicine Institutional Review Board approval. 2-and 4-Step Scales. This study addressed a standard 2-class classification problem-referable or nonreferable AMD-that was based on the original AREDS 4-step scale; details and criteria for the scale are described in the Box. Grades 3 and 4 in the 4-step scale correspond to higher risk for progression to advanced AMD.

Deep learning in ophthalmology: The technical and clinical considerations

“…In addition, OCT angiography may be able to identify additional features such as deep capillary plexus nonperfusion associated with macular photoreceptor damage 217,218 . This is compatible with studies showing that the long-term recovery of photoreceptor integrity and visual outcome in DME is dependent on perfusion status of the deep capillary plexus 219 223 , severity characterization and estimation of 5-year risk 95 and disease conversion 224 . In addition, a severity classification based on fundus photography was developed 16 .…”

Next-Generation Sequencing Applications for Inherited Retinal Diseases

“…This involves expert curation to obtain the training dataset but has the advantage of removing variability due to being a single assessor once training is complete. For example, many studies employ ML to assess fundus and optical coherence tomography images to train and predict outcomes for potential age-related macular degeneration (AMD) patients, including risk of progression to a more severe disease state and response to treatments [ 162 , 163 , 164 , 165 , 166 , 167 , 168 , 169 ]. This approach has enormous potential benefits for improvement of diagnostic and prognostic accuracy.…”

A Comparison of Transfer Learning Performance Versus Health Experts in Disease Diagnosis From Medical Imaging

“…Burlina et al, [22] were applied to the 5664 color fundus images obtained from the NIH AREDS dataset to detect outer boundaries of the retina and resize them to 231x231 to conform the overFeat network. The National Eye Institute Age-Related Eye Disease Study (AREDS) dataset used by Burline et al, [56] which has total 67401 number of fundus photograph and it is identified as a gold-standard dataset. The open access series of breast ultrasonic dataset, which contains 882 images of unique breast masses, consists of 678 benign and 204 malignant lesions [23].…”