Medical Record Synthesis using GPT-3.5

Author Dr.Nyein Chan Ko Ko (2023)

Overview

This project leverages GPT-3.5 to synthesize medical records, aiming to provide a compliant solution for researchers needing access to patient data while respecting privacy laws like HIPAA and GDPR.

Repository Contents

• main.py: Main function execution
• data_generation.py: Synthetic patient profile and record generation
• gpt_processing.py: GPT model utilization for record creation
• file_management.py: Transformation of GPT texts into CSV
• required_libraries.py: Library imports

Objectives

• Evaluate synthetic medical record generation feasibility and accuracy
• Process analysis from prompt engineering to database-ready notes
• Apply findings in language model research and prototyping

Methodology

• Prompt Engineering: Develop prompts for GPT-3.5 to accurately generate medical data
• Data Generation: Create patient profiles and detailed medical records
• Data Processing: Convert generated texts to structured CSV format

Observations

• Generated 1079 balanced patient records by nationality and gender
• Noted realistic age distributions and disease prevalences
• Addressed potential biases and the importance of reliable data sources

Applications

• Enhancing NLP tasks in healthcare
• Supporting medical education and research
• Prototype testing within privacy regulations

Conclusion

The project highlights the potential of GPT-3.5 in synthesizing realistic medical records, providing a valuable tool for medical research within ethical and legal standards.

Usage

Install dependencies listed in required_libraries.py. Run main.py to begin synthetic medical record generation. Follow component-specific instructions for detailed operations.

Read more about this project here