OphtAI: Artificial Intelligence for Diabetic Retinopathy screening

Only for diabetic retinopathy, we would need to be able to analyze 2,3 million images per day to fullfill WHO-recommended annual screening for diabetic patients. In this context, Artificial Intelligence is an essential and fundamental asset in view of the many challenges to be solved.

Diabetic Retinopathy Computer-Aided Diagnosis Service

  • Medical Device (EC Class IIa MD) 
  • Result of several years of collaborative research work between ADCIS, Evolucare, Mines ParisTech, INSERM and AP-HP (Paris University Hospitals)
  • Works with one or more photos of the retina per eye or per patient
  • Compatible with most fundus cameras:fixed or handheld, DICOM or not


  • Image quality assessment
  • Determination of the existence and assessment of the severity of Diabetic Retinopathy
  • Suggested course of action: to refer patient to an ophthalmologist in case of more than mild retinopathy
  • Lesion mapping superimposed on retina photo:
    • aneurysms
    • haemorrhages
    • cotton-wool nodules
    • exudates
    • neovessels
Performance characteristics 


  • Response time: <3s
  • Sensitivity of 99%
  • Specificity of 87% 
Sensitivity: proportion of correct results in patients with Diabetic Retinopathy
Specificity: proportion of correct results in patients without Diabetic Retinopathy

Artificial Intelligence powered by Ophtai: learn more on www.ophtai.com