NoseCheck - DNS Screening Tool

A Python-based system for detecting nasal asymmetry using smartphone photos and symptom assessment.

🔗 Try it live → nosecheckai-v2.onrender.com

Project Overview

NoseCheck is a research project that aims to detect Deviated Nasal Septum (DNS) using:

1. Computer Vision Analysis: Automated facial landmark detection and nasal asymmetry measurements
2. Symptom Assessment: Self-reported symptom questionnaire
3. Calibration System: Validated against 3D-printed models with known deviations

📸 How It Works

Upload a frontal face photo → answer a short symptom questionnaire → get a deviation score with severity classification (Normal / Mild / Moderate / Severe).

Photo Upload → MediaPipe Face Landmarks → Asymmetry Metrics → Deviation Score → Result

Features

• Automated facial landmark detection using MediaPipe
• Quantitative nasal asymmetry measurements
• Multi-metric deviation scoring system
• Symptom questionnaire integration
• Calibration with 3D-printed reference models
• Statistical analysis and visualization
• Mobile-friendly web interface

Project Structure

nosecheck/
├── src/
│   ├── image_processing/      # Photo capture and preprocessing
│   ├── landmark_detection/    # Facial landmark detection using MediaPipe
│   ├── measurement/           # Nasal asymmetry calculations
│   ├── scoring/              # Deviation score algorithm
│   ├── questionnaire/        # Symptom checklist module
│   ├── data_management/      # Data storage and retrieval
│   └── analysis/             # Statistical analysis and visualization
├── data/
│   ├── calibration_models/   # Known deviations of 3D models
│   ├── images/               # Captured photos
│   └── results/              # Measurement data and scores
├── models/                   # MediaPipe model files (face_landmarker.task)
├── notebooks/                # Jupyter notebooks for analysis
├── tests/                    # Unit tests
├── docs/                     # Documentation and research notes
└── scripts/                  # Utility scripts

Installation

Prerequisites

• Python 3.8+ (Python 3.9.6+ recommended)
• pip (Python package manager)

Step-by-Step Setup

1. Clone or download this repository

   git clone https://github.com/lakshman1213/nosecheck.git
   cd nosecheck

2. Create a virtual environment

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   For full development (includes pandas, matplotlib, scikit-learn for analysis):

   pip install -r requirements.txt

   For minimal web app only (production-like):

   pip install -r requirements-production.txt

4. Download the MediaPipe Face Landmarker model

   The model will auto-download on first run, but you can download it manually:

   mkdir -p models
   curl -L -o models/face_landmarker.task \
     "https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task"

Running Locally

After installation, start the Flask web server:

Option 1: Using the convenience script

python run_server.py

Option 2: Using Python module

python -m src.app

Option 3: Using Flask CLI

python -m flask --app src.app run --host 0.0.0.0 --port 5001

The server will start on http://localhost:5001

• Open your browser and navigate to http://localhost:5001
• The web interface provides a mobile-friendly UI for photo upload, symptom questionnaire, and results display
• Press Ctrl+C to stop the server

Development vs Production Dependencies

• requirements.txt: Full development dependencies including pandas, matplotlib, scikit-learn, jupyter. Use this for local development, calibration scripts, and data analysis.
• requirements-production.txt: Minimal dependencies for web app only. Use this for production deployments or if you only need the web interface.

Quick Start

Web Interface (Recommended)

1. Start the server: python run_server.py
2. Open http://localhost:5001 in your browser
3. Upload a frontal face photo
4. Complete the symptom questionnaire
5. View your screening result with deviation score and classification

Python API

from src.landmark_detection.detector import detect_landmarks
from src.measurement.asymmetry_calculator import calculate
from src.scoring.score_calculator import calculate_score

# Process an image
image_path = "path/to/nose/photo.jpg"
landmarks = detect_landmarks(image_path)
if landmarks:
    measurements = calculate(landmarks)
    result = calculate_score(measurements)
    print(f"Deviation Score: {result['deviation_score']}")
    print(f"Classification: {result['classification']}")

Key Metrics

The system calculates the following nasal asymmetry metrics:

1. Lateral Deviation: Distance from nose tip to facial midline
2. Septal Angle: Angle of nasal septum from vertical
3. Nostril Asymmetry: Left vs. right nostril size/shape differences
4. Bridge Alignment: Straightness of nasal bridge

Technology Stack

• Computer Vision: OpenCV, MediaPipe
• Data Processing: NumPy, Pandas
• Statistical Analysis: SciPy, scikit-learn
• Visualization: Matplotlib, Seaborn
• Web Interface: Flask
• Database: SQLite (via SQLAlchemy)
• Development: Jupyter, pytest, black

Development

Running Tests

pytest tests/

Code Formatting

black src/

Troubleshooting

Common Issues

Port 5000 already in use (macOS)

• macOS AirPlay Receiver uses port 5000 by default
• NoseCheck uses port 5001 to avoid conflicts
• If port 5001 is also in use, change it in run_server.py or src/app/__main__.py

Model download fails

• Check your internet connection
• Download manually:

  mkdir -p models
  curl -L -o models/face_landmarker.task \
    "https://storage.googleapis.com/mediapipe-models/face_landmarker/face_landmarker/float16/1/face_landmarker.task"

Import errors

• Ensure virtual environment is activated: source venv/bin/activate
• Reinstall dependencies: pip install -r requirements.txt
• Check Python version: python3 --version (should be 3.8+)

Face detection fails

• Ensure photo is frontal (face camera directly)
• Check lighting (even, not too dark or overexposed)
• Photo should show full face with nose clearly visible
• Try a different photo with better quality

Module not found errors

• Make sure you're in the project root directory
• Verify src/ directory exists and contains all modules
• Check that __init__.py files exist in each package directory

Research Validation

This tool is designed for research purposes to:

• Validate correlation between calculated scores and known deviations
• Assess repeatability of measurements
• Evaluate classification accuracy across severity categories
• Serve as an educational tool for adolescent health awareness

Expected Outcomes

• Quantitative Validation: High correlation (r²) between known and calculated deviations
• Repeatability: Coefficient of variation < 10% for repeated measurements
• Classification Accuracy: Reliable distinction between normal/mild/moderate/severe categories

Contributing

This is a research project. For questions or collaboration inquiries, please contact the research team.

License

[To be determined based on research institution requirements]

Disclaimer

This tool is for research and educational purposes only. It is not intended for clinical diagnosis or medical decision-making. Always consult qualified healthcare professionals for medical advice.

Citation

[To be added upon publication]