AI system screens for referrable DM-associated retinopathy

  • Natarajan S & al.
  • JAMA Ophthalmol
  • 8 Aug 2019

  • International Clinical Digest
Access to the full content of this site is available only to registered healthcare professionals. Access to the full content of this site is available only to registered healthcare professionals.

Takeaway

  • An offline artificial intelligence (AI) algorithm can detect referable diabetic retinopathy ([R]DR) images taken from a smartphone-based portable camera.

Why this matters

  • Current ophthalmology-based screening methods are costly and time-consuming and require training.

Study design

  • Prospective study of the Medios AI (Remidio) automated system.
  • System analyzes retinal images (3 fundus fields) taken with a smartphone-based nonmydriatic retinal camera.
  • A minimally trained health care worker imaged 213 patients with diabetes mellitus (DM) at clinics in Mumbai, India.
  • RDR: retinopathy more severe than mild DR, with/without macular edema.
  • Results compared with direct ophthalmologist grading.
  • Funding: Medios Technologies provided the AI software.

Key results

  • Of 187 patients diagnosed as not having RDR by ophthalmologists, AI correctly diagnosed 92%.  
  • Of 15 with ophthalmologist-diagnosed RDR, AI correctly diagnosed 100%.
  • Of 12 with ophthalmologist-diagnosed mild nonproliferative DR, AI diagnosed 67%.
  • AI sensitivity and specificity (95% CIs) were, respectively:
    • RDR: 100% (78.2%-100%) and 88.4% (83.16%-92.53%); and 
    • Any DR: 85.2% (66.3%-95.8%) and 92.0% (87.1%-95.4%).

Limitations

  • Small sample size.
  • Current AI version does not allow grading according to International Clinical DR severity scale or National Health Service classification.
  • Ungradable images resulting from cataracts and small pupil size necessitated dilated retinal photography screening.
  • Local setting, generalizability uncertain.  

Please confirm your acceptance

To gain full access to GPnotebook please confirm:

By submitting here you confirm that you have accepted Terms of Use and Privacy Policy of GPnotebook.

Submit