Automated Extraction of Unstructured Post-SBRT Toxicity Data from Radiology Reports Using Large Language Models

TL;DR

This study demonstrates that large language models can accurately extract specific toxicity and progression outcomes from unstructured radiology reports in lung SBRT patients, streamlining clinical data extraction.

Contribution

We developed prompt-engineered LLM methods to reliably extract toxicity and progression data from unstructured radiology reports, showing high accuracy and viability for clinical applications.

Findings

High sensitivity and specificity in toxicity detection

Effective classification of progression status

Viability of LLMs for clinical report extraction

Abstract

We evaluated the viability of using a Large Language Model (LLM) to extract patient-specific specific toxicity and progression outcomes from unstructured radiology reports. We retrospectively extracted 160 follow-up CT and PET/CT electronic medical record notes for patients treated with lung stereotactic body radiotherapy (SBRT) at our institution from January 2017 through December 2023. Using the Llama 3.3-70-B-Instruct LLM, we engineered prompts to extract four clinical endpoints from each radiology report: locoregional progression, distant progression, radiation-induced fibrosis, and radiation-induced rib fractures. Progression endpoints were classified as yes, no, or maybe, while fibrosis and rib fractures were binary (yes or no). Ground truth labels were defined using two-grader consensus for the 60-note training set, used for prompt development, and a three-grader majority vote for the 100-note test set. LLM performance was evaluated using sensitivity, specificity, and accuracy. As detailed by our evaluation metrics, the strong performance of our methods demonstrates the viability of using prompt-engineered LLMs to extract radiation-toxicities and progression classification from radiology reports.