Award
NIH Reporter #11448840
Automated PET Report Generation with Visual Grounding
Recipient
University of Wisconsin-Madison
Award Amount
$595,791.00
Ceiling
$595,791.00
Awarded
September 09, 2026
Identifier
11448840
Development of an automated PET report generation framework using federated learning and visual grounding to improve report turnaround time and quality, addressing workforce shortages amid increasing PET scan volumes.
Description
This project aims to develop a novel framework for automated positron emission tomography (PET) report generation, addressing the growing imbalance between the limited workforce and increasing PET scan volumes due to the growth of theranostics. The framework will use federated learning to overcome PET’s data scarcity problem, and will employ visual grounding techniques and longitudinal reasoning to create personalized, interactive PET reports. The ultimate goal is to expedite report turnaround time while maintaining or improving report quality and clarity, potentially leading to faster diagnoses and improved patient care.