Analysis and optimization of Large Language Model-based methods for information extraction from radiology reports

This project therefore focuses on optimizing Large Language Models for extracting structured information from radiology reports. The primary objectives include creating a baseline model pipeline for clinical information extraction based on the linguistic theory of frame semantics and maximizing model performance while considering limited hardware and data resources. Furthermore, human-grounded evaluation of model explainability is performed, and the comparability and reproducibility of studies describing LLM-based information extraction should be enhanced.