Mining information from clinical texts by looping the experts in the training process

This project focuses on developing an advanced NLP framework to unlock insights from unstructured clinical text within Electronic Health Records (EHR). Leveraging over 20 million clinical texts at Humanitas Research Hospital, the initiative addresses challenges in the Italian healthcare context, including less structured data and limited language-specific tools. Using active learning and Large Language Models (LLMs), the framework minimizes manual annotation while actively engaging clinical experts, ensuring both technical accuracy and clinical relevance. Automating tasks like outcome retrieval, which currently demand extensive manual effort, will save time and reduce clinician workload. By integrating these tools into EHR systems, this project aims to enhance decision-making, streamline clinical research, and support personalized, efficient care pathways, advancing sustainable healthcare practices.