Diabetic retinopathy classification for supervised machine learning algorithms

Luis Filipe Nakayama,Lucas Zago Ribeiro,Mariana Batista Gonçalves,Paulo Henrique Morales,Helen Nazareth Veloso Dos Santos,Caio Vinicius Saito Regatieri,Mauricio Maia,Daniel A Ferraz,Fernando Korn Malerbi,Rubens Belfort Mattos

doi:10.1186/s40942-021-00352-2

Abstract

BackgroundArtificial intelligence and automated technology were first reported more than 70 years ago and nowadays provide unprecedented diagnostic accuracy, screening capacity, risk stratification, and workflow optimization.Diabetic retinopathy is an important cause of preventable blindness worldwide, and artificial intelligence technology provides precocious diagnosis, monitoring, and guide treatment. High-quality exams are fundamental in supervised artificial intelligence algorithms, but the lack of ground truth standards in retinal exams datasets is a problem.Main bodyIn this article, ETDRS, NHS, ICDR, SDGS diabetic retinopathy grading, and manual annotation are described and compared in publicly available datasets. The various DR labeling systems generate a fundamental problem for AI datasets. Possible solutions are standardization of DR classification and direct retinal-finding identifications.ConclusionReliable labeling methods also need to be considered in datasets with more trustworthy labeling.

Highlights

This study compared the most often-applied Diabetic retinopathy (DR) classification scales: Scottish Diabetic Retinopathy Grading [14], Early Treatment Diabetic Retinopathy Grading [15], International Clinic Diabetic Retinopathy [16], National Health Service Diabetic Retinopathy Classification grading [17], Modified Davis Retinopathy staging [18], and direct findings identification.The Early Treatment Diabetic Retinopathy Study At an international consortium of ophthalmologists at Airlie House in 1968, internists and neurosurgeons standardized a diabetic retinopathy classification applied in the landmark Early Treatment Diabetic Retinopathy Study [15], designed to generate a more precise staging for DR and macular edema
Blots, linear hemorrhages, and microaneurysms were classified as red spots when they were not distinguished in Early Treatment Diabetic Retinopathy Grading (ETDRS) charts [19]
Artificial intelligence and automated technology were first reported more than 70 years ago and nowadays provide unprecedented diagnostic accuracy, screening, risk stratification, and workflow optimization [3]

Summary

Background

Computers executing automated functions were first described in 1950, with the first publication in 1943. Automated technology provides unprecedented diagnostic accuracy, screening capacity, risk stratification, and workflow optimization with accuracy equivalent to healthcare professionals [3] and more cost-effective diseases screening [4]. Diabetic retinopathy (DR) is the leading cause of preventable blindness in working-age adults worldwide [7, 8], responsible for more than 24,000 annual cases of blindness [9] and the main focus in Ophthalmological AI screening algorithms [10]. There is an increased blindness risk in patients with chronic diabetes mellitus, especially those with poor clinical control [11]. Telemedicine and automated screening programs could diagnose, monitor, and guide treatment. This article describes and compares the most commonly diabetic retinopathy classifications, referencing criteria, and their applications in datasets

Main text

Findings

Conclusions