Suraj Eye Institute is building AI models for diabetic retinopathy, glaucoma, AMD, myopia, and ocular biometry — trained on India’s most clinically rich Central Indian retinal dataset. Developed in Nagpur. Validated for India.
Vidarbha has over 23 million people. Ophthalmologist-to-patient ratio: less than 1:100,000 in rural areas. AI can bridge that gap — but only if it is trained and validated on this population.
Most AI systems for eye disease are trained primarily on datasets from Western populations, or major South Indian hospitals. They are not validated on the Central Indian population — which has distinct genetic risk factors, disease severity patterns, and demographic profiles.
SEI sees 100+ patients daily across two campuses. Every patient is a data point. With structured collection and ethical governance, SEI can build the largest labelled ophthalmic dataset from Central India — and use it to develop AI that actually works for this population.
This is not aspirational. It starts today, with existing equipment, existing staff, and free open-source tools.
Each model is trained on the best publicly available global datasets and SEI’s own growing Fundus Bank, then progressively fine-tuned for superior performance on the Central Indian population.
India has the world’s second-largest diabetic population. In Vidarbha, diabetic retinopathy (DR) is a leading cause of avoidable blindness — most cases caught too late. Our AI grades DR severity automatically from a fundus photograph.
Glaucoma is the leading cause of irreversible blindness worldwide — and is largely asymptomatic until late stages. Our model analyses fundus images for optic nerve head changes, cup-to-disc ratio, and RNFL patterns.
Age-related macular degeneration and diabetic macular oedema are detected on OCT — but interpretation requires specialist expertise. Our AI analyses OCT scans for drusen, fluid, and structural changes indicating macular disease.
Myopia is a growing global epidemic, particularly among younger populations. Our AI model analyses fundus photographs and biometric data to screen for myopia and predict progression risk — enabling early intervention in school-based and community screening programmes.
AI-assisted analysis of ocular biometric measurements — axial length, corneal curvature, anterior chamber depth, and lens thickness — to establish Central Indian normative data, improve IOL power calculations, and support refractive surgery planning.
Starting today — at zero additional cost — every patient at SEI gets a labelled fundus photograph. Every OCT scan is stored in a structured, diagnosis-tagged archive. This is the foundation that future AI is built on.
Every tool in our AI development stack is open-source or freely available. This is by design — SEI’s AI programme is a proof that cutting-edge medical AI does not require expensive enterprise platforms.
Industry-standard deep learning framework used by Google, Meta, and top research labs worldwide.
FREE · Open SourceState-of-the-art convolutional neural networks, pre-trained on ImageNet and fine-tuned for retinal images.
FREE · Transfer LearningFree T4 GPU compute for model training. Kaggle Notebooks provide additional 30 hrs/week free GPU.
FREE · Cloud GPUBuild a web interface for the AI model in hours — upload a fundus image, get a prediction and confidence score.
FREE · Open SourceEyePACS, MESSIDOR-2, ORIGA, DRISHTI-GS, ACRIMA, OCTDL, RFMiD — 90,000+ training images, all free.
FREE · Open AccessVersion control, collaboration, and open-source publication of SEI’s AI code — building international credibility.
FREE · Open SourceA phased roadmap from raw data to validated, published AI — designed to be completed with existing SEI resources and open-source tools.
Systematic labelling protocol established. Public datasets downloaded and organised. All optometrists and technicians briefed. Collection begins at both campuses.
Training a 5-class DR grading model on EyePACS (88,702 images) using EfficientNet on Google Colab. Target: >90% accuracy on standard benchmark.
Training glaucoma detection model using ORIGA + DRISHTI-GS + ACRIMA datasets. Cup-to-disc ratio analysis and optic nerve head classification.
Training AMD and DME detection model on OCTDL + RFMiD datasets. OCT scan analysis for drusen, fluid, and macular structural changes.
Building a Gradio web app — upload a fundus image, get instant AI grading with confidence scores. This is the live demo shown to potential partners and at conferences.
As the Fundus Bank reaches 5,000+ labelled images, models are retrained on local data. Performance on Central Indian patients improves substantially vs. global-only training.
AI vs. clinician on 500 consecutive cases at SEI. IRB-approved study design. This is the data needed for publication and regulatory submission.
“Development and Validation of an In-House AI Model for Glaucoma Detection at a Central Indian Eye Hospital.” Target journals: British Journal of Ophthalmology, IOVS, Indian Journal of Ophthalmology.
When SEI publishes a validated AI model for eye disease detection — built and tested in Central India — it will be the first of its kind from a non-AIIMS, non-Sankara Nethralaya, non-LV Prasad hospital. It puts SEI on the AI ophthalmology map permanently.
Whether you are an AI company seeking validation, a researcher seeking data, a funder seeking impact, or a clinical partner seeking credibility — there is a place for you in this programme.
Bring your AI model. SEI provides prospective clinical validation on a diverse Central Indian patient cohort, with co-authored publication and regulatory-ready evidence.
Access to SEI’s labelled Fundus Bank and the Central India Ophthalmology Biobank for academic and commercial research, under a formal data sharing agreement.
Partner with SEI’s clinical team and AI developers to co-build models trained specifically on the Central Indian population — and publish the results together.
Fund the acquisition of portable OCT for mobile screening, computing resources, or laboratory infrastructure through CSR (Schedule VII compliant) or research grants.
That is SEI’s philosophy on AI. We are not waiting for perfection. We are building, validating, publishing — and improving from there. Join us.
You cannot copy content of this page