Poster 246

by in  Poster Session 3

How is Plus Disease Diagnosed in ROP? Insights from a Deep Learning Computer-Based Image Analysis System with Occlusion Analysis

Layla M. Ghergherehchi, MD; James M. Brown, PhD; Susan Ostmo, MS; Sang Jin Kim, MD, PhD; John P. Campbell, MD; R.V. Paul Chan, MD; Jayashree Kalpathy-Cramer, PhD; Michael F. Chiang, MD
Casey Eye Institute, Oregon Health & Science University
Portland, OR


Introduction: Diagnosis of plus disease in retinopathy of prematurity (ROP) is subjective and variable.1 We have shown that computer-based image analysis can diagnose plus disease with comparable or better accuracy than experts.2 This study uses outputs from a deep learning algorithm to identify vascular features considered most significant by experts for plus disease diagnosis.

Methods: 31 wide-angle retinal images with plus disease, based on a consensus reference standard, were selected for this study. Occlusion analysis was performed using a convolutional neural network (CNN) to compute the significance of each 12×12 pixel image region to the network’s ability to make a diagnosis, visualized as a ‘heat map’. Vascular features were extracted from the areas that most relatively increased or decreased the probability of diagnosis.

Results: Retinal features identified as being important for plus disease based on occlusion analysis were: (1) central retinal location of vessels (31/31 images), (2) mid-peripheral location of vessels (25/31 images), (3) arterial tortuosity (31/31 images), (4) venous dilation (31/31 images), (5) arterial dilation (31/31), (6) venous tortuosity (31/31).

Discussion: Experts are often unable to explain their diagnostic process,3 and occlusion analysis methods can provide important insight about this process. Study findings show that many features considered important for diagnosis are not included in the definition of plus disease.

Conclusion: Vascular abnormalities including dilation and tortuosity of both arteries and vein in all fields of view are important for the diagnosis of retinopathy of prematurity. This has important implications for clinical care and education in ROP diagnosis.

References: 1. Chiang MF, Jiang L, Gelman R, Du YE, Flynn JT. Interexpert agreement of plus disease diagnosis in retinopathy of prematurity. Arch Ophthalmol 2007;125(7):875-80.
2.         Hewing NJ, Kaufman DR, Chan RV, Chiang MF. Plus disease in retinopathy of prematurity: qualitative analysis of diagnostic process by experts. JAMA Ophthalmol 2013;131:1026-32.
3.         Kalpathy-Cramer J, Campbell JP, Kim SJ, et al. Deep learning for the identification of plus disease in retinopathy of prematurity. Invest. Ophthalmol. Vis. Sci. 2017;58:5554.

Print Friendly, PDF & Email

Leave a Comment

Your email address will not be published. Required fields are marked *

Comment *