Researchers compare large language models against traditional ontology methods for extracting breast cancer data from unstructured medical notes

LLM-based framework developed to automatically extract breast cancer phenotypes from unstructured oncology provider notes in Electronic Medical Records

System extracts critical clinical information including chemotherapy outcomes, biomarkers, tumor location, size, and growth patterns that oncologists document in natural language

Study compares the new LLM approach against knowledge-driven annotation systems using NCIt Ontology Annotator to evaluate extraction accuracy

Research addresses real-world EMR challenges where oncologists prefer entering clinical insights as natural language text rather than using structured data fields