Reproduce “PhenoFlow: A Human-LLM Driven Visual Analytics System for Exploring Large and Complex Stroke Datasets” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jaeyoung Kim, Sihyeon Lee, Hyeon Jeon, Keon-Joo Lee, Hee-Joon Bae, Bohyoung Kim, Jinwook Seo
Researchers from Seoul National University and collaborating hospitals developed PhenoFlow, a visual analytics system that employs a Large Language Model as a "data wrangler" to efficiently construct and explore patient cohorts within extensive stroke datasets. This system empowered neurologists to efficiently analyze data, reduce cognitive load, and uncover previously unrecognized clinical patterns, like U-shaped and triangular blood pressure trajectories, while ensuring patient data privacy.
Key insights
- Integrating LLMs as "data wranglers" significantly reduces the cognitive burden on domain experts by automating complex data preparation and query tasks, allowing them to focus on clinical reasoning.
- Trust in AI-driven tools can be established through a transparent human-in-the-loop workflow, where LLM-generated executable code and visual inspection views allow experts to verify and debug the AI's output.
- Novel visualization techniques, like the slice-and-wrap design, can effectively reveal subtle yet clinically relevant temporal patterns in dense, irregularly sampled time-series medical data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.