India has one of the world’s largest diabetes burdens, with an estimated 101 million people living with diabetes, representing an overall weighted prevalence of 11.4%. Diabetic retinopathy (DR), including diabetic macular edema (DME), is an important cause of vision impairment and blindness among people with diabetes. Early detection and timely treatment can help prevent irreversible vision loss.
The Challenge
Screening for both DR and DME currently relies on retinal imaging and expert review by trained ophthalmologists—resource-intensive and difficult to scale uniformly, especially with a rural ophthalmologist-to-population ratio of 1:219,000.
Our AI Solution
MadhuNetrAI is an AI-enabled screening solution designed to assist the health workforce- optometrists, Community Health Officers (CHOs) and medical officers- in early identification of DR and DME from fundus images.
MadhuNetrAI expands access to diabetic eye screening by enabling DR and DME assessment through portable retinal imaging and AI, including in settings where specialist availability is limited.
By providing standardized, data-driven screening insights, it supports timely referral and helps healthcare providers prioritize patients requiring specialist attention.
The solution demonstrates high diagnostic performance, achieving 96.9% sensitivity and 94.7% specificity, with a Cohen’s kappa of 0.91.
How it Works
01
Preparatory Stage
The patient is informed about the procedure and moved to an environment suitable for fundus imaging.
Preparatory Stage
The patient is informed about the procedure and moved to an environment suitable for fundus imaging.
02
Fundus Imaging
A clear retinal image is captured using a tabletop or handheld portable fundus camera.
Fundus Imaging
A clear retinal image is captured using a tabletop or handheld portable fundus camera.
03
Image Upload and Pre-Processing
Images are uploaded to MadhuNetrAI, which pre-processes them to flag duplication and ensure quality output.
Image Upload and Pre-Processing
Images are uploaded to MadhuNetrAI, which pre-processes them to flag duplication and ensure quality output.
04
AI Assessment
The model assigns a DR grade based on severity — Grade 0 (No DR), Grade 1 (Mild), Grade 2 (Moderate), Grade 3 (Severe), Grade 4 (Proliferative) and presence or absence of DME.
AI Assessment
The model assigns a DR grade based on severity — Grade 0 (No DR), Grade 1 (Mild), Grade 2 (Moderate), Grade 3 (Severe), Grade 4 (Proliferative) and presence or absence of DME.
05
Follow-Up with Experts
The patient is counseled for follow-up or referral to an appropriate specialist for further evaluation and treatment.
Follow-Up with Experts
The patient is counseled for follow-up or referral to an appropriate specialist for further evaluation and treatment.
Where It's Deployed
MadhuNetrAI is deployed across 47 facilities pan-India.
13
States
Attribution and Impact
- Identification of At-Risk Individuals: Since implementation, MadhuNetrAI has identified ten percent of screened patients with mild DR and detected DME in 18 percent including 44 cases among patients with mild DR.
- Reduced Risk: The solution has referred nearly about 38 percent of those screened with moderate to proliferative DR and/or presence of DME for specialist intervention, thereby reducing the risk of vision-threatening progression.
- Efficient Resource Optimization: The AI solution helps optometrists independently conduct DR screenings, thereby allowing ophthalmologists to focus on more complex cases.
- Economic Impact and Scalability: Early AI-enabled screening helps avert increased cost of care per patient, strengthening the case for scaling up the solution within public health systems.
Our Other Solutions
Partners
Looking Ahead
Quality healthcare is the foundation of a healthy society. Sustained partnerships and continuous collaboration will help refine the technology and broaden its reach into new domains of ophthalmic care.