Synthetic Data for Medical Imaging AI
 A collage image of cups with barcodes, a scientist looking in microscope, virus, and bottle of medication.
Title: Synthetic Data for Medical Imagining AI
Date: November 14, 2024
Time: 12:00 p.m. - 1:00 p.m. ET
Webcast
Speaker:
Elena Sizikova, Ph.D.
Staff Fellow
Office of Science and Engineering Laboratories (OSEL)
Center for Devices and Radiological Health (CDRH)
U.S. Food and Drug Administration (FDA)
About the Speaker:
Elena Sizikova is a staff fellow in the Office of Science and Engineering Laboratories, Center for Devices and Radiological Health (CDRH) in the Food and Drug Administration (FDA) in Silver Spring, MD. Her work addresses research problems at the intersection of artificial intelligence (AI), medical imaging and computer vision, focusing on the use of synthetic data in regulatory science.
About the Presentation:
Artificial Intelligence (AI)-enabled medical imaging devices require access to large-scale and representative datasets for both training and evaluation. Obtaining sufficient data remains a crucial challenge for most applications in medical image analysis, in part due to patient privacy concerns, acquisition and annotation difficulties or high costs, limiting wider development of medical AI. We show that synthetic data, i.e., artificial data designed to approximate properties and relationships seen in patient data, can be used to supplement patient data in medical AI development and evaluation, mitigating data availability concerns. We summarize and compare different methodologies for creating medical synthetic data, distinguishing between knowledge-based (KB) (e.g., mechanistic) and imaging-based (e.g., generative AI) models. We demonstrate how KB synthetic data generation in breast and skin imaging applications can be effectively used for AI development and analysis.
www.FDA.gov/GrandRounds
|
