Automated Diagnosis of Plus Disease in Retinopathy of Prematurity using Deep Learning
Peter Campbell; James Brown; RV Paul Chan; Jennifer Dy; Stratis Ioannidis; Deniz Erdogmus; Jayashree Kalpathy-Cramer; Michael F. Chiang
Oregon Health & Science University
Introduction: The diagnosis of plus disease is highly variable and there is growing evidence this translates into real world treatment differences between clinicians. This paper presents and evaluates a fully automated algorithm based on deep learning (DL) for diagnosis of plus disease in ROP from retinal images and proposes a real world solution to this problem.
Methods: We developed a convolutional (deep) neural network (i-ROP DL) to diagnose three level plus disease using a multi-institutional database of nearly 6000 Retcam images with reference standard diagnosis (consensus of ophthalmoscopy and three or more telemedicine diagnoses). Performance was evaluated using area under receiver operating characteristic curves (AUC) and comparison with diagnoses from eight independent clinical experts using quadratic weighted kappa coefficients. We are developing a web-based centralized server with open source access to enable this technology to be utilized anywhere with an internet connection.
Results: Mean AUCs of 0.94±0.01 for diagnosis of pre-plus or worse disease and 0.98±0.01 for diagnosis of “plus” (versus not plus) were observed. The algorithm outperformed 6 out of 8 experts in the test set, demonstrating a quadratic weighted kappa score of 0.92 compared to the RSD, whereas the mean kappa among 8 experts compared to the RSD was 0.85 (range 0.80 – 0.95).
Discussion: The i-ROP DL algorithm can diagnose plus disease automatically with the same or higher proficiency than ROP experts. We will present a fully automated, open source screening platform for incorporation into telemedicine programs.
Conclusion: The i-ROP DL system has potential to improve the quality, accessibility, and cost of ROP screening worldwide.
1. Fleck BW, Williams C, Juszczak E, et al. An international comparison of retinopathy of prematurity grading performance within the Benefits of Oxygen Saturation Targeting II trials. Eye (Lond). 2017;123:1-7. doi:10.1038/eye.2017.150.
2. Campbell JP, Ataer-Cansizoglu E, Bolon-Canedo V, et al. Expert Diagnosis of Plus Disease in Retinopathy of Prematurity From Computer-Based Image Analysis. JAMA Ophthalmol. 2016;134(6):651-657. doi:10.1001/jamaophthalmol.2016.0611.
3. Wittenberg LA, Jonsson NJ, Paul Chan RV, Chiang MF. Computer-Based Image Analysis for Plus Disease Diagnosis in Retinopathy of Prematurity. J Pediatr Ophthalmol Strabismus. 2011;49(1):11-19. doi:10.3928/01913913-20110222-01.