Application of Small Language Models in Structured Extraction of Pediatric Pathology Reports: An Efficient and Privacy-Preserving Medical Information Processing Solution

Introduction / Main Post: Application of Small Language Models in Structured Extraction of Pediatric Pathology Reports: An Efficient and Privacy-Preserving Medical Information Processing Solution

This article introduces a semi-automatic annotation workflow based on small language models, which can extract structured information from unstructured pediatric histopathology reports in a CPU-only environment, achieving an accuracy rate of 84.3% while protecting patient privacy.