Abstract

Modern healthcare generates an ocean of real-world data (RWD) — from electronic health records and clinical notes to medical imagings and insurance claims. These data capture the full spectrum of patient experiences outside clinical trials, offering invaluable insights for improving care and advancing research. Yet, their diversity, unstructured nature, and quality variability often turn them into an untapped resource rather than a reliable evidence base.

At Regenstrief Institute and Indiana University Health, decades of data integration have created one of the nation’s richest RWD ecosystems, encompassing longitudinal EHR data, clinical narratives, and medical images. The question now is: How can we make this wealth of data truly “speak” to us?

This seminar explores how large language models (LLMs) are beginning to answer that question — transforming how we extract, interpret, and utilize real-world clinical information. We will walk through four interconnected stories:

1. Information Extraction – How LLMs unlock knowledge buried in unstructured clinical notes.

2. Medical Coding Assistance – How domain-specific fine-tuned LLMs enhance coding accuracy and efficiency.

3. Patient–Trial Matching – How natural language understanding accelerates recruitment by linking patients to the right clinical studies.

4. From Image to Report – How vision–language models are automating lung cancer screening interpretation and report generation.

By the end of this talk, you will see how RWD and LLMs together are not just improving efficiency, but reimagining what learning from everyday clinical practice can mean for the future of medicine.

Biography

Dr. He’s research focuses on leveraging innovative informatics methodologies to advance healthcare and biomedical research. His work emphasizes data standardization and integration, real-world data-driven frameworks for clinical trial eligibility criteria, clinical decision support systems, and cohort discovery platforms powered by artificial intelligence, including large language models. Dr. He has also developed health informatics tools, including platforms for knowledge graph visualization, ontology annotation systems, and semantic data integration frameworks. Additionally, he applies real-world data to create computable phenotypes, driving advancements in eHealth and mHealth, and uses natural language processing (NLP) techniques to enhance data extraction and integration.

Dr. He’s interdisciplinary work bridges data science and medicine, enabling transformative solutions that improve patient care and population health. His contributions drive innovation in health informatics, expanding the applicability and impact of advanced data-driven methodologies in healthcare systems.

tudents registered for the seminar are expected to attend in-person.

Teams ID and Passcode:

Meeting ID: 211 123 896 292 8

Passcode: Uh9qs2pf