Past Event: Babuška Forum

Dynamic Imaging of Tumor Vascular Perfusion using 4D Photoacoustic Computed Tomography

Umberto Villa, Research Scientist, Oden Institute for Computational Engineering and Sciences, University of Texas at Austin

Abstract

The ability to perform dynamic imaging of time-varying physiological processes in small animal models is critically needed to understand the progression of human disease and develop new therapies. Although dynamic imaging methods have been used to evaluate tumor vascular perfusion in small animal models, the available methods typically provide only two-dimensional (2D) spatial imaging, lack the precision needed for quantitative measurements, or suffer from other drawbacks. Photoacoustic computed tomography (PACT) can circumvent the limitations of existing methods and has been recognized as a promising tool for dynamic small animal imaging. By exploiting the optical absorption of hemoglobin or exogenous contrast agents, dynamic PACT holds great potential for measuring important time-varying biomarkers such as tumor vascular perfusion and oxygenation and improving the assessments of anti-cancer and other therapies. However, the data acquisition design of commercially available PACT imagers poses significant challenges for dynamic imaging because only a few tomographic views are available to reconstruct each temporal frame. As such, the image reconstruction problem is severely ill-posed. In this work, I will present two different spatiotemporal image reconstruction methods to address this challenge by exploiting redundancies in the sought-after time-varying image. These approaches use low-rank matrix estimation and neural field representations to construct an image prior that not only mitigates the ill-posedness of the reconstruction problem but also reduces the computational and memory burden. Numerical studies demonstrate that the proposed methods accurately estimate the temporal evolution of a contrast agent in a tumor model by use of an anatomically realistic numerical mouse phantom.

This is a joint work with Luke Lozenski (WashU), Refik Cam (UIUC), and Dr. Mark Anastasio (UIUC).

Biography

Dr. Villa obtained his Ph.D. in Mathematics from Emory University in 2012 with a specialization in Computational Mathematics (high-performance computing, computational fluid dynamics, image processing, inverse problems, and numerical analysis), with Dr. Alessandro Veneziani serving as advisor. After completing his postdoctoral training at the Center for Applied Scientific Computing (CASC) of Lawrence Livermore National Laboratory (LLNL), Dr. Villa joined the Oden Institute as a Research Associate, working with Dr. Omar Ghattas on scalable numerical methods for the solution of Bayesian inverse problems, uncertainty quantification and propagation, optimal experimental design, and optimization under uncertainty. In 2022, after four years as a Research Assistant Professor of Electrical & Systems Engineering at Washington University in St Louis, Dr. Villa rejoined the Oden Institute, where, as a Research Scientist, is developing a research agenda at the interface of imaging science and predictive scientific computing.