Project Grant 2205265

This four-year, $1.1 million Project Grant from the National Science Foundation's Computer and Information Science and Engineering program will fund the development of physics-informed machine learning techniques to analyze dynamic blood flow from static subtraction computed tomographic angiography imaging. The University of Wisconsin-Milwaukee will train graduate students in deep learning methods and engage undergraduates and local high school students, particularly those from underrepresented communities, in biomedical engineering research. The researchers will develop neural network models of contrast concentration, three-dimensional blood velocity, and relative pressure to infer dynamic hemodynamic data from routinely collected medical images. They will validate the methods using numerical simulations and in vitro flow experiments. The University of Utah will receive a sub-award of an unspecified amount to create 3D transport models, synthetic imaging data, and tools to quantify wall shear stress. By enabling extraction of additional hemodynamic information from existing medical scans, this research aims to advance personalized cardiovascular treatment without requiring new imaging hardware or protocols.