Rajpurkar et al. (2021)
paperAuthors
Credibility Rating
Gold standard. Rigorous peer review, high editorial standards, and strong institutional reputation.
Rating inherited from publication venue: Nature
Demonstrates deep learning system for medical image analysis achieving specialist-level performance on retinal disease detection, relevant to AI safety through validation of healthcare AI reliability, robustness across datasets, and clinical deployment considerations.
Paper Details
Metadata
Summary
Rajpurkar et al. (2021) developed a deep learning platform (DLP) capable of detecting 39 different fundus diseases and conditions from retinal photographs using 249,620 labeled images. The system achieved high performance metrics (F1 score of 0.923, sensitivity of 0.978, specificity of 0.996, AUC of 0.9984) on multi-label classification tasks, reaching the average performance level of retina specialists. External validation across multiple hospitals and public datasets demonstrated the platform's effectiveness, suggesting potential for retinal disease triage and screening in remote areas with limited access to ophthalmologists.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| AI-Human Hybrid Systems | Approach | 91.0 |
Cached Content Preview
Automatic detection of 39 fundus diseases and conditions in retinal photographs using deep neural networks | Nature Communications
Skip to main content
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
Advertisement
Automatic detection of 39 fundus diseases and conditions in retinal photographs using deep neural networks
Download PDF
Download PDF
Subjects
Diagnostic markers
Machine learning
Retinal diseases
Abstract
Retinal fundus diseases can lead to irreversible visual impairment without timely diagnoses and appropriate treatments. Single disease-based deep learning algorithms had been developed for the detection of diabetic retinopathy, age-related macular degeneration, and glaucoma. Here, we developed a deep learning platform (DLP) capable of detecting multiple common referable fundus diseases and conditions (39 classes) by using 249,620 fundus images marked with 275,543 labels from heterogenous sources. Our DLP achieved a frequency-weighted average F1 score of 0.923, sensitivity of 0.978, specificity of 0.996 and area under the receiver operating characteristic curve (AUC) of 0.9984 for multi-label classification in the primary test dataset and reached the average level of retina specialists. External multihospital test, public data test and tele-reading application also showed high efficiency for multiple retinal diseases and conditions detection. These results indicate that our DLP can be applied for retinal fundus disease triage, especially in remote areas around the world.
Similar content being viewed by others
Intelligent retinal disease detection using deep learning
Article
Open access
08 December 2025
Detection of signs of disease in external photographs of the eyes via deep learning
Article
29 March 2022
Deep learning-based automated detection fo
... (truncated, 68 KB total)3bea36a1b8cd3738 | Stable ID: sid_t9nAh9RFe7