Anemia Screening Chatbot

A multi-modal anemia risk screening system combining questionnaire-based assessment with optional eye image analysis using deep learning.

📸 Application Screenshots

Landing Page

Modern, professional landing page with medical disclaimers

Chat Interface

Interactive chatbot with progressive questioning

Results Dashboard

Comprehensive risk assessment with AI-powered analysis

Screening History

Track all past screenings with detailed statistics

Note: Screenshots will display once images are added to the screenshots/ folder.

---

🎯 Live Demo

Run locally to see the application in action:

python app.py
# Open http://localhost:5000

Key Features:

• 💬 Interactive chat interface with smooth animations
• 🤖 ML-powered image analysis (EfficientNet-B0)
• 📊 Comprehensive risk dashboard
• 📈 Screening history tracking
• 🎨 Professional medical-grade UI

---

⚠️ Medical Disclaimer

This tool is for screening purposes only and does NOT provide medical diagnosis.

• Not a substitute for professional medical advice
• Not a replacement for laboratory testing
• Always consult healthcare providers for proper diagnosis

🎯 Features

• Interactive Chatbot Interface: Natural conversation flow for symptom assessment
• Rule-Based Screening: Evidence-based risk scoring from clinical indicators
• Optional Image Analysis: Deep learning analysis of conjunctival images (EfficientNet-B0)
• Multi-Modal Fusion: Combines questionnaire + image data for improved accuracy
• Instant Results Dashboard: Clear risk visualization and recommendations
• Session History: Track screening results over time

🏗️ Architecture

Frontend (HTML/CSS/JS)
    ↓
Flask Backend (app.py)
    ↓
├─→ Rule Engine (rules.py) ──→ Risk Score
├─→ ML Model (ml_model.py) ──→ Image Analysis
└─→ Database (database.py) ──→ Storage
    ↓
Results Dashboard

📁 Project Structure

anemia_chatbot/
│
├── app.py                 # Flask backend
├── rules.py              # Risk calculation logic
├── ml_model.py           # ML model integration
├── database.py           # SQLite database
├── requirements.txt      # Dependencies
│
├── templates/
│   ├── index.html       # Landing page
│   ├── chat.html        # Chat interface
│   └── dashboard.html   # Results display
│
├── static/
│   ├── style.css        # Styling
│   └── chat.js          # Frontend logic
│
└── uploads/             # User images (created automatically)

🚀 Quick Start

1. Install Dependencies

cd anemia_chatbot
pip install -r requirements.txt

2. Run the Application

python app.py

3. Open Browser

Navigate to: http://localhost:5000

📊 Model Integration

The system uses a trained EfficientNet-B0 model from the parent directory:

../models/image2/efficientnet_b0_cv3_models.pt

Model Performance:

• Accuracy: 66.0%
• ROC-AUC: 0.712
• Trained on 468 pediatric conjunctival images

If the model is not found, the system falls back to rule-based screening only.

🔧 Configuration

Database

• SQLite database: anemia_screening.db (auto-created)
• Stores: answers, risk levels, scores, image paths

Image Upload

• Max file size: 16MB
• Accepted formats: PNG, JPG, JPEG
• Storage: uploads/ folder

Risk Scoring

Rule-Based Points:

• Age < 18: +1
• Female gender: +1
• Each symptom (fatigue, dizziness, etc.): +1-2
• ML prediction high (>0.7): +3
• ML prediction moderate (0.5-0.7): +2
• ML prediction low-moderate (0.3-0.5): +1

Risk Levels:

• Low Risk: 0-3 points
• Moderate Risk: 4-8 points
• High Risk: 9+ points

🎨 User Flow

1. Landing Page → User clicks "Start Screening"
2. Chat Interface → Bot asks 10 questions (7 required, 3 optional)
3. Image Upload → Optional eye image upload
4. Processing → Rule-based + ML analysis
5. Dashboard → Results, recommendations, risk score

🧪 Testing

Test Without ML Model

# System works with rules only if model file is missing
python app.py

Test With ML Model

# Ensure model path is correct in ml_model.py
# Default: ../models/image2/efficientnet_b0_cv3_models.pt
python app.py

📈 API Endpoints

GET  /               Landing page
GET  /chat           Chat interface
POST /api/submit_answer    Save answer
POST /api/upload_image     Upload eye image
POST /api/calculate_risk   Get risk assessment
GET  /dashboard      Results page
GET  /api/history    Get screening history

🔒 Privacy & Security

• No personal identifiers collected
• Images stored locally with timestamps
• Session-based data handling
• No external API calls
• SQLite database (local storage)

🛠️ Development

Add New Questions

Edit rules.py → get_questions() function

Modify Risk Logic

Edit rules.py → calculate_risk() function

Update ML Model

Replace model file and update path in ml_model.py

Customize UI

Edit files in templates/ and static/ folders

📝 Citation

If using this system for research:

Pediatric Anemia Detection using Multi-Modal AI
Authors: [Your Name]
Year: 2025
Institution: [Your Institution]

🐛 Troubleshooting

"Model not found" warning:

• Check model path in ml_model.py
• System works without model (rule-based only)

Images not uploading:

• Check uploads/ folder permissions
• Verify file size < 16MB

Database errors:

• Delete anemia_screening.db to reset
• Database auto-recreates on restart

📚 Requirements

• Python 3.8+
• Flask 3.0+
• PyTorch 2.1+
• Pillow (PIL)
• NumPy

🎓 Educational Use

This project demonstrates:

• Full-stack web development
• ML model deployment
• Medical AI ethics
• User experience design
• Multi-modal data fusion

📧 Support

For questions or issues, please refer to the code comments or create an issue in the repository.

---

Remember: This is a screening tool, NOT a diagnostic system. Always consult healthcare professionals for medical advice.