Automatic Extraction of Structured Information from Brain MRI Reports Using an Open-Weight Large Language Model

Researchers successfully used LLaMA 3.1 to extract structured medical data from Dutch brain MRI reports, achieving high accuracy on visual rating scores and diagnostic mentions. The study demonstrates strong potential for automating medical data extraction from clinical reports using open-source models. Few-shot prompting techniques significantly improved performance on numerical variables, suggesting practical pathways for clinical deployment.