Fri, Jun 12, 2026, 08:00 – 09:00AM (PDT), 11:00 – 12:00 noon (EST)
Title: Statistics in the Age of Generative AI: Reconstructing, Resolving, and Synthesizing Clinical Evidence
Abstract:
Clinical evidence synthesis increasingly requires methods that can integrate information from heterogeneous and often incomplete sources, including published survival curves, reconstructed individual patient data, subgroup summaries, and free-text trial eligibility criteria. At the same time, the rapid development of generative AI creates new opportunities to extract, structure, and analyze evidence that was previously difficult to use at scale. In this talk, Dr. Xu will present a research program at the intersection of generative AI and statistical inference for modern clinical evidence synthesis.
Three connected projects will be discussed. First, KM-GPT uses multimodal AI, image processing, and iterative reconstruction algorithms to automate the recovery of individual patient-level survival data from published Kaplan-Meier plots. Second, RESOLVE-IPD advances this task by developing statistically principled methods for high-fidelity IPD reconstruction, including approaches that use censoring information and address incomplete subgroup reporting. Third, EligMeta extends evidence synthesis beyond reconstructed survival data by using AI agents to transform clinical questions and trial eligibility criteria into structured, auditable workflows for eligibility-aware meta-analysis. Together, these projects show that generative AI can serve as an interface between unstructured clinical evidence and rigorous statistical inference. Rather than replacing statistical reasoning, AI can help convert figures, text, and incomplete summaries into analyzable statistical objects, while statistical methods provide the structure needed for reproducibility, uncertainty quantification, and valid inference.
Speaker: Dr. Yanyan Xu
Dr. Yanxun Xu is a Professor and the Joseph & Suzanne Jenniches Faculty Scholar in the Department of Applied Mathematics and Statistics, the Data Science and AI Institute, and the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University. Her research lies at the intersection of Bayesian statistics, artificial intelligence, and biomedical data science, with methodological contributions to reinforcement learning, high-dimensional data analysis, nonparametric statistics, and uncertainty quantification. Her work has been applied broadly across intelligent healthcare, including clinical trial design, cancer genomics, early disease diagnosis, predictive modeling for Alzheimer’s disease, and the analysis of electronic health records. She is a Fellow of the American Statistical Association.
