Umberto Villa
Assistant Professor,
Department of Biomedical Engineering
Oden Institute for Computational Engineering and Sciences
The University of Texas at Austin
I work at the interface of imaging science and predictive scientific computing — developing reconstruction algorithms for emerging biomedical imaging modalities, and the methods needed to evaluate them rigorously.
About
My research combines engineering, mathematical modeling, artificial intelligence, and high-performance computing to advance quantitative biomedical imaging. The current focus is on emerging acoustic and optical modalities — photoacoustic computed tomography (PACT) and ultrasound computed tomography (USCT) — with the goal of advancing the state of care for cancer diagnosis and treatment.
I obtained my Ph.D. in Mathematics from Emory University in December 2012, specializing in computational mathematics, with Prof. Alessandro Veneziani as my principal advisor. I completed my postdoctoral training at the Center for Applied Scientific Computing of Lawrence Livermore National Laboratory (2013–2015), working with Dr. Panayot Vassilevski on algebraic multigrid and numerical upscaling for flow in porous media.
I then joined the Oden Institute at UT Austin (2015–2018) as a Research Associate, working with Prof. Omar Ghattas on scalable numerical methods for Bayesian inverse problems, uncertainty quantification, optimal experimental design, and optimization under uncertainty. After four years as a Research Assistant Professor of Electrical and Systems Engineering at Washington University in St. Louis and a member of the Imaging Science PhD faculty, I rejoined the Oden Institute in August 2022.
Research areas
-
Task-based image quality
Numerical observers and information-theoretic bounds that measure what an imaging system lets a clinician actually do.
Read more -
Trustworthy deep learning
Exposing the failure modes of learned reconstruction — hallucination, miscalibrated uncertainty — and designing methods that can be analyzed.
Read more -
Virtual imaging trials
Stochastic numerical phantoms and end-to-end simulation pipelines, with the ground truth and sample sizes that credible evaluation demands.
Read more -
Physics-based reconstruction
Accurate imaging operators and novel object representations for photoacoustic and ultrasound computed tomography.
Read more -
Inverse problems & UQ
Scalable algorithms for infinite-dimensional Bayesian inversion, delivered as sustainable open-source software.
Read more -
Open-source software
hIPPYlib, hIPPYlib-MUQ, SOUPy, ParELAG — tools used by the wider computational science community.
Read more
Recent publications
- AI Based Cough Analysis for Pulmonary Tuberculosis Triage and Diagnosis: A Technical Review
- A virtual imaging framework for three-dimensional quantitative optoacoustic tomography using stochastic numerical Breast phantoms
- Taylor approximation variance reduction for approximation errors in PDE-constrained Bayesian inverse problems
- A Convergent Generalized Krylov Subspace Method for Compressed Sensing MRI Reconstruction with Gradient-Driven Denoisers
- A Learning-based Framework for Spatial Impulse Response Compensation in 3D Photoacoustic Computed Tomography
We are looking for undergraduate, MS, and PhD students to join the lab. See open positions for how to get in touch.