Large Language Models as a Rapid and Objective Tool for Pathology Report Data Extraction

Area of Science:

• Medical Informatics
• Digital Pathology
• Artificial Intelligence in Medicine

Background:

• Pathology departments generate vast amounts of data crucial for scientific research.
• Manual data extraction from pathology reports is labor-intensive, time-consuming, and prone to errors.
• Developing efficient data extraction methods is essential for creating large-scale research databases.

Purpose of the Study:

• To evaluate the effectiveness of large language models (LLMs) in extracting data from pathology reports.
• To compare the performance of different LLMs (ChatGPT and Google Bard) in converting unstructured pathology data into a synoptic format.
• To assess the potential of AI-assisted data extraction for enhancing academic research.

Main Methods:

• Ten de-identified pathology reports from resection specimens were selected.
• Reports were processed using ChatGPT and Google Bard.
• Both AI models were tasked with converting reports into a synoptic format suitable for data editors like Excel or Google Sheets.

Main Results:

• Both LLMs successfully created tabular data from the pathology reports.
• Google Bard demonstrated an advantage by automatically facilitating spreadsheet creation.
• The AI-assisted approach showed promise in overcoming manual data extraction challenges.

Conclusions:

• Large language models offer a viable solution for automating data extraction in pathology research.
• AI-assisted data extraction can significantly improve efficiency and precision compared to manual methods.
• Implementing LLMs can accelerate the creation of valuable research databases from existing pathology archives.