This research demonstrates how Large Language Models (LLMs) can efficiently extract structured information from unstructured breast cancer histopathology reports without requiring labeled training data.

• Achieves extraction through zero-shot prompting with natural language instructions
• Evaluates LLM performance across multiple medical data extraction tasks
• Enables scalable automation of a traditionally manual, time-consuming process
• Addresses data privacy considerations for sensitive medical information

This advancement matters for healthcare by making valuable clinical data more accessible for research while reducing manual effort, potentially accelerating discoveries and improving patient outcomes.

Leveraging large language models for structured information extraction from pathology reports