Skip to content
Longterm Wiki
Back

De Fauw et al. (2018)

paper

Credibility Rating

5/5
Gold(5)

Gold standard. Rigorous peer review, high editorial standards, and strong institutional reputation.

Rating inherited from publication venue: Nature

A landmark Nature Medicine paper illustrating how interpretability and human-AI collaboration concerns manifest in high-stakes medical AI deployment; often cited in discussions of safe AI integration into clinical workflows.

Metadata

Importance: 62/100journal articleprimary source

Summary

De Fauw et al. present a deep learning system that diagnoses over 50 retinal diseases from OCT scans with expert-level accuracy by separating segmentation and classification into two sequential neural networks. The system achieves performance matching or exceeding world-leading retinal specialists and provides interpretable, clinically actionable referral recommendations. This work demonstrates both the promise and the interpretability challenges of deploying AI in high-stakes medical decision-making.

Key Points

  • Two-stage architecture: first network segments retinal anatomy into 3D maps, second network classifies disease from those maps, improving interpretability.
  • Achieved expert-level diagnostic performance across 50+ retinal conditions, matching or surpassing world-leading ophthalmologists.
  • The segmentation-first approach allows clinicians to verify intermediate representations, partially addressing the 'black box' concern in medical AI.
  • Demonstrates generalization across different OCT scanner types, a key practical challenge for clinical deployment.
  • Raises important questions about human-AI collaboration, appropriate trust calibration, and responsibility in automated medical referrals.

Cited by 1 page

PageTypeQuality
AI-Human Hybrid SystemsApproach91.0

Cached Content Preview

HTTP 200Fetched Mar 15, 202630 KB
Clinically applicable deep learning for diagnosis and referral in retinal disease | Nature Medicine 
 
 
 

 
 

 
 

 

 
 
 
 

 

 
 
 
 
 
 

 
 
 
 
 
 

 
 

 
 
 
 
 
 
 
 
 
 
 

 
 

 

 

 
 

 
 
 

 
 

 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 

 
 
 

 
 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

 
 
 
 
 
 
 
 
 
 
 

 
 
 
 

 

 
 
 
 

 

 

 
 
 
 
 
 
 
 
 

 
 
 Subjects

 
 Diagnosis 
 Eye manifestations 
 Machine learning 
 Three-dimensional imaging 

 
 

 
 
 

 
 

 
 

 
 Abstract

 The volume and complexity of diagnostic imaging is increasing at a pace faster than the availability of human expertise to interpret it. Artificial intelligence has shown great promise in classifying two-dimensional photographs of some common diseases and typically relies on databases of millions of annotated images. Until now, the challenge of reaching the performance of expert clinicians in a real-world clinical pathway with three-dimensional diagnostic scans has remained unsolved. Here, we apply a novel deep learning architecture to a clinically heterogeneous set of three-dimensional optical coherence tomography scans from patients referred to a major eye hospital. We demonstrate performance in making a referral recommendation that reaches or exceeds that of experts on a range of sight-threatening retinal diseases after training on only 14,884 scans. Moreover, we demonstrate that the tissue segmentations produced by our architecture act as a device-independent representation; referral accuracy is maintained when using tissue segmentations from a different type of device. Our work removes previous barriers to wider clinical use without prohibitive training data requirements across multiple pathologies in a real-world setting.

 

 
 
 
 
 
 
 
 
 
 Access through your institution 
 
 
 
 
 
 
 
 Buy or subscribe 
 
 
 
 
 
 

 
 
 

 
 
 
 
 
 
 This is a preview of subscription content, access via your institution 

 
 
 

 

 Access options

 

 
 
 
 
 
 
 
 Access through your institution 
 
 
 
 
 
 

 

 
 
 
 
 
 Access Nature and 54 other Nature Portfolio journals
 

 
 Get Nature+, our best-value online-access subscription
 

 
 
 $32.99 / 30 days 
 

 cancel any time

 
 
 Learn mor

... (truncated, 30 KB total)
Resource ID: 0068f706474a5ab0 | Stable ID: MTZkYzczMT