Detecting sight-threatening eye disease earlier
We are developing and validating artificial-intelligence tools for the eye diseases that cause the most avoidable sight loss in Central India — so that, in future, more people can be screened earlier and closer to home.
AI that is validated for Central India
Most AI systems for eye disease have been trained on images from other populations. Differences in genetics, disease severity and imaging equipment can change how well they perform. Our programme develops and tests AI with the patients it is meant to serve — before any clinical use is considered.
A large, under-served region
Vidarbha is home to more than 23 million people, many of whom live far from an eye specialist.
Specialist-labelled images
Retinal photographs and OCT scans graded by experienced clinicians at two campuses.
Ethics first
Consent, de-identification and Institutional Ethics Committee oversight at every step.
Five conditions, one goal: earlier diagnosis
Diabetic retinopathy
Automated grading of diabetic retinopathy severity from retinal photographs — from no retinopathy to proliferative disease.
Model developmentGlaucoma
Assessment of the optic nerve head and cup-to-disc ratio on retinal photographs, to flag possible glaucoma for specialist review.
In developmentMacular disease on OCT
Detection of drusen, fluid and structural change in age-related macular degeneration and diabetic macular oedema.
In developmentMyopia
Screening for short-sightedness and estimating the risk of progression from retinal images and biometry — for school and community programmes.
Data collectionOcular biometry & lens power
Population-specific analysis of axial length, corneal curvature and anterior-chamber depth to improve intraocular-lens power calculation in cataract surgery.
Data collectionWhat comes next
Each model is evaluated against specialist diagnosis before it is considered for clinical use.
Our approach →The foundation: a carefully curated image archive
Good AI depends on good data. Project Fundus Bank is a structured, consented collection of retinal images built to the standards that research and regulators expect.
- LabelledRetinal photographs and OCT scans are labelled with the diagnosis by the examining clinician.
- OrganisedImages are filed by condition — glaucoma, diabetic retinopathy, AMD, normal and other.
- LinkedEach image is linked to clinical information in our electronic medical record.
- ProtectedAll data are de-identified before research use; patient privacy is protected throughout.
From data to clinical use — step by step
- CurateBuild a high-quality, specialist-labelled image collection.
- DevelopTrain models on established international benchmark datasets.
- AdaptRefine them with de-identified images from Central Indian patients.
- ValidateCompare AI with specialist diagnosis in a prospective, ethics-approved study.
- PublishShare the results in peer-reviewed journals.
- ImplementConsider clinical use only after successful validation and regulatory approval.
Collaborate on AI in eye care
We work with AI developers seeking clinical validation, researchers seeking collaboration, and public-health partners interested in community screening. We would be glad to hear from you.