73P SCRIPT: Stratified clinical risk prediction from pathology reports using large language models

Abstract

Accurate risk stratification in oncology is essential for guiding treatment decisions, but is limited by the complexity and unstructured format of healthcare data. Pathology reports are mostly composed of unstructured text. These reports contain rich prognostic information, which cannot be easily used for quantitative risk prediction models. Here, we use Large language models (LLMs) to extract diverse risk factors from these reports and yield a single quantitative survival risk score.

Ähnliche Arbeiten