June 9 -10 | Boston

AI, Large Language Models and Data Science Workflows in i2b2

June 9

Welcome: Symposium overview Diane Keogh

Keynote Introduction: Gilbert “Gil” Omenn, MD, PhD.

Keynote: Human Values Project: The AI we want for the Research we need. Zak Kohane, MD, PhD.

The Future of i2b2 and LLMs: Transforming Cohort Discovery and Beyond? 

Moderator: Zak Kohane, MD, PhD.
Panel: Nathan Palmer, PhD, Sandy Aronson, ALM, MA, Shawn Murphy, MD, PhD

Computational Phenotypes – Federated Learning.

Griffin Weber, MD, PhD

Slides

Tianxi Cai, ScD

Slides

Sponsor: Chartis

Networks Panel Discussion

  • ENACT – Enclaves Use Cases – Shyam Visweswaran, MD, PhD.
  • MassCPR – Griffin Weber, MD. PhD.
  • Federated Learning – Shawn Murphy, MD, PhD.

Enabling Data Science workflows supported by new features in i2b2 1.8.2

  • Data Export Demonstration – Jeff Klann, PhD.
  • *Coming soon* New and Improved i2b2 Foundation website – Nich Wattanasin
  • Docker Containers – Kavi Wagholikar, MBBS, PhD.

Working group updates
● UI Working group/Committee on Technology – Griffin Weber, MD, PhD
● Ontology Working Groups – Michele Morris

Ontology to support additional data domains in i2b2 – Abu Mosa, PhD, MS, FAMIA.

Questions and Answers

Virtual attendees questions answered.

Debate: Federated vs Centralized Networks
Optimizing Medical Data for Research: Centralized, Federated, and Hybrid Approaches

Moderator: Shawn Murphy

Panelists: Abu Mosa, PhD; Jeff Klann, PhD; Thomas H. McCoy, MD.

June 10

Machine Learning in i2b2 to address data quality issues: Loyalty Cohorts/Computational Phenotypes, CIPHER Demo.

Jeff Klann, PhD, Griffin Weber, MD, PhD,  Jackie Honerlaw, RN, MPH

With its initial launch twenty years ago, i2b2 enabled wide scale access to clinical data for research. This workshop will take a deep dive into how i2b2 is now using machine learning algorithms to address data quality issues to generate more accurate and trustworthy results. Attendees will learn the steps of implementing these algorithms within their own i2b2 system and see a demo of the Centralized Interactive Phenomics Resource (CIPHER), where they can explore computational phenotype models generated by other institutions.

Griffin Weber, MD, PhD.

Jeff Klann, PhD.

Jackie Honerlaw, RN, MPH.

Hands on: Using LLM to Search for Patient Notes
This training module is focused on Large Language Models (LLMs) and their application to clinical informatics and patient note analysis.  Using Jupyter notebooks that demonstrate Retrieval-Augmented Generation (RAG) semantic search, embeddings, and structured clinical summarization using local LLMs

Shawn Murphy, MD, PhD, Valdery Moura Junior, PhD, MBA 

Curating precision PASC Research Cohort for clinical studies on Long Covid

This workshop will introduce an open-source algorithm designed to identify patient-specific conditions following COVID-19, improving the precision of PASC (Long-COVID) research. Attendees will gain hands-on experience using synthetic data to apply the algorithm, explore its broader potential for detecting other chronic conditions, and learn how to integrate AI-driven diagnostics into their own research using data from i2b2 instances.

 

Hossein Estiri, PhD and Jonas Hügel, PhD.