Mid-Project Presentation: AI-Powered Mobile App for Early Cataract Detection
The project focuses on developing an AI-powered mobile application for early cataract detection, with a special focus on making vision care more accessible in rural and underserved areas.
The Problem
In many remote villages, people must travel long distances just to get an eye check-up. This consumes time, money, and energy, and often delays treatment. Without timely detection, cataracts can worsen and even lead to blindness. Additionally, rural areas often lack well-equipped facilities or trained personnel.
The Solution
The team’s mobile app offers a simple, affordable, and AI-driven solution. With just a smartphone camera, community health workers can capture eye images and receive real-time predictions about the presence of cataracts. This helps avoid unnecessary hospital visits and ensures early intervention.
Objectives
- Collect and organize medical image data effectively.
- Build a user-friendly mobile app with AI-powered predictions.
- Collaborate with doctors to test the model, gain feedback, and improve accuracy.
Understanding Cataracts
A cataract is the clouding of the eye’s natural lens. It causes blurry vision, glare, and faded colors, and can lead to blindness if untreated. The team studied cataracts across stages (early, immature, mature) to build the dataset needed for AI training.
The Technology
One challenge with AI models is dataset bias—models may learn irrelevant details like skin tone instead of eye features. To solve this, the team used Meta’s SAM 2 (Segment Anything Model) to precisely isolate the pupil. Their custom model, SAM CAT, is a lightweight CNN (Convolutional Neural Network), efficient enough for low-resource rural environments.
App System Design
The app includes:
- Registration/Login using Aadhaar, phone number, and basic details.
- Profile Management to store patient records.
- Database Screen with previous scans and history.
- Eye Image Capture & Edit with options to upload or take new photos, crop, and center the pupil.
- AI Prediction Screen showing cataract detection results.
- Visit Records to track patient history over time.
Collaboration and Learning
The team collaborated actively using VS Code Live Share, Google Meet, and WhatsApp discussions. Their learning journey included:
- Business problem-solving and aligning solutions to real needs.
- App design, database management, and UI prototyping.
- AI model development, prompt engineering, and Python frameworks.
- Team communication, collaboration, and presentation skills.
Future Goals
Looking ahead, the team plans to:
- Complete integration of the database feature.
- Add advanced pupil extraction tools.
- Incorporate the AI model for live predictions.
- Test, debug, and refine the app before deployment.