Seminar, Hosted by National Cancer Institute, Center for Biomedical Informatics and Information Technology (CBIIT)
Introducing Imaging Data Commons Agentic Interfaces
Brigham and Women's Hospital and Harvard Medical School
Abstract
NCI's Imaging Data Commons (IDC) holds more than 100 terabytes of freely available multimodal imaging from clinical and basic cancer research studies. This session introduces two new tools, enabled by large language models, that let researchers search, explore and work with IDC data by describing their needs in everyday language: the IDC Model Context Protocol (MCP) server and an agent skill that integrates with LLM-enabled assistants. The demonstration covers setting up the MCP server and agent skill to connect an assistant to IDC, finding and accessing DICOM imaging data sets and cohorts from natural-language descriptions of a research question, and using the assistant to brainstorm research questions or hypotheses for a specific data set, followed by live Q&A. The IDC is one of seven data commons in NCI's Cancer Research Data Commons.