Leveraging large language models for structured information extraction from pathology reports

Structured information extraction from unstructured histopathology reports facilitates data accessibility for clinical research. Manual extraction by experts is time-consuming and expensive, limiting scalability. Large language models (LLMs) offer efficient automated extraction through zero-shot prompting, requiring only natural language instructions without labeled data or training. We evaluate LLMs' accuracy in extracting structured information from breast cancer histopathology reports, compared to manual extraction by a trained human annotator.

We developed the Medical Report Information Extractor, a web application leveraging LLMs for automated extraction. We also developed a gold-standard extraction dataset to evaluate the human annotator alongside five LLMs including GPT-4o, a leading proprietary model, and the Llama 3 model family, which allows self-hosting for data privacy. Our assessment involved 111 breast cancer histopathology reports from the Generations study, extracting 51 pathology features specified within the study's data dictionary.

Evaluation against the gold-standard dataset showed that both Llama 3.1 405B (94.7% accuracy) and GPT-4o (96.1%) achieved extraction accuracy comparable to the human annotator (95.4%; p = 0.146 and p = 0.106, respectively). Whereas Llama 3.1 70B (91.6%) performed below human accuracy (p < 0.001), its reduced computational requirements make it a viable option for self-hosting.

We developed an open-source tool for structured information extraction that demonstrated expert human-level accuracy in our evaluation using state-of-the-art LLMs. The tool can be customized by non-programmers using natural language and the modular design enables reuse for diverse extraction tasks to produce standardized, structured data facilitating analytics through improved accessibility and interoperability.