🚑 GLiner-TransbronchialBiopsy

A specialized medical Named Entity Recognition (NER) system for analyzing transbronchial biopsy reports, powered by fine-tuned GLiNER models.

🎯 Project Overview

GLiner-TransbronchialBiopsy is designed specifically for extracting medical entities from transbronchial biopsy reports, with a focus on transplant rejection analysis. The system combines state-of-the-art NLP techniques with domain-specific medical knowledge.

🔍 Key Features

• Specialized Medical NER: Tailored for transbronchial biopsy reports
• Interactive Annotation: Real-time medical text processing
• Comprehensive Entity Coverage: Focuses on critical biopsy parameters
• Performance Optimization: Fine-tuned for medical domain accuracy

🔧 Technical Requirements

• Python 3.9+
• CUDA-compatible GPU (recommended)
• 8GB RAM minimum
• 2GB free disk space

📦 Installation

# Create and activate virtual environment
python -m venv gliner-env
source gliner-env/bin/activate  # Unix/macOS
gliner-env\Scripts\activate     # Windows

# Install dependencies
pip install -r requirements.txt

🎓 Supported Medical Entities

Entity Type	Description	Example
Site	Biopsy location	"LSD", "LM"
Fragment Count	Total fragments analyzed	"4 fragments"
Alveolar Count	Number of alveolar fragments	"3 fragments alvéolés"
Rejection Grade	A/B grading scale	"Grade A2"
Chronic Rejection	Presence indicators	"Rejet chronique minimal"
C4d Staining	Staining results	"C4d négatif"

and others ...

💻 Usage Guide

Model Training

from gliner_transbronchial import TrainingConfig

config = TrainingConfig(
    data_path="./data/biopsy_reports.json",
    output_dir="./models/production",
    batch_size=8,
    learning_rate=2e-5,
    num_epochs=10
)

trainer.train(config)

Interactive Dashboard

streamlit run app.py --server.port 8501

🔄 Development Workflow

1. Data Preparation

   • Report collection
   • Manual annotation
   • Quality assurance

2. Model Training

   • Hyperparameter optimization
   • Cross-validation
   • Error analysis

3. Evaluation

   • Performance metrics
   • Clinical validation
   • Error analysis

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push to branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.