Down Syndrome Detection Mobile App

The Problem

Early detection of Down syndrome is crucial for timely intervention and better outcomes for affected children. Traditional screening methods often require specialized medical equipment and trained personnel, making them less accessible in many communities.

The Solution

I built a mobile app that brings Down syndrome screening directly to parents and healthcare workers through their smartphones. The app uses the device’s camera to capture facial images and runs ML analysis locally, providing immediate screening results without needing internet connectivity.

How It Works

The app captures facial landmarks using Dlib’s 68-point facial landmark detector, then extracts features using Local Binary Patterns (LBP). These features are fed into a Support Vector Machine (SVM) classifier trained on a dataset of facial images from children with and without Down syndrome.

Technical Details

• Frontend: Flutter for cross-platform mobile development
• ML Framework: Dlib for facial landmark detection
• Feature Extraction: Local Binary Patterns (LBP)
• Classification: Support Vector Machine (SVM)
• Target Age: Children aged 0-6 years

Results

The model achieved 93% accuracy and 98% AUC on test data, demonstrating strong potential as a preliminary screening tool. While not a replacement for medical diagnosis, it can help identify children who may benefit from further medical evaluation.

Challenges & Learnings

Working with facial recognition for children presented unique challenges — kids move, their facial features are still developing, and capturing consistent images was difficult. I learned a lot about data augmentation and the importance of diverse training datasets.