Facial Nerve Paralysis Severity Grading by Computer Vision and Machine Learning
Jul 26, 2022·,,,·
1 min read
Bingnan Huo
Keith Buffinton
Joshua Stough
Arun K Gadre
Facial Nerve Paralysis Grading SystemAbstract
This research project developed an automated system for diagnosing and grading the severity of Facial Nerve Paralysis (FNP) using computer vision and machine learning techniques. The system achieved over 80% accuracy in grading FNP severity by processing facial images. We implemented image processing techniques to standardize unstructured data and trained multiple machine learning models including Random Forest, SVM, and XGBoost, with comprehensive cross-validation for performance evaluation.
Type
Publication
Susquehanna Valley Undergraduate Research Symposium
Overview
This undergraduate research project focused on developing an automated system for diagnosing Facial Nerve Paralysis (FNP) using computer vision and machine learning techniques. The work resulted in a Python-based backend that achieved over 80% accuracy in grading FNP severity.
Key Contributions:
- Developed image processing pipeline for standardizing unstructured facial image data
- Implemented data augmentation techniques for improved model training
- Trained and evaluated multiple machine learning models (Random Forest, SVM, XGBoost)
- Created a comprehensive cross-validation framework for performance evaluation
Technologies Used:
- Python
- Pandas & NumPy for data processing
- OpenCV for image processing
- Matplotlib & Seaborn for visualization
- scikit-learn for machine learning models
The project was awarded First Place Best Poster at the 2022 Susquehanna Valley Undergraduate Research Symposium.
Repository Structure
The codebase is available on GitHub and includes:
- Image processing pipeline
- Feature extraction methods
- Model training and evaluation scripts
- Cross-validation framework