Division of Nuclear Medicine, Department of Radiology
Statistical signal and image processing, medical imaging, nuclear imaging, tomographic imaging, nonparametric estimation, parametric estimation, estimation bounds, optimization algorithms, imaging system design, robust estimation
The research in my group is focused on developing and analyzing model-based methods for signal and image processing problems. Our particular focus is on tomographic image reconstruction based on accurate physical and statistical models of medical imaging systems, and related inverse problems such as image restoration. There are four main thrusts of the research that my students and I perform.
1) Developing new iterative algorithms for inverse problems. These algorithms are tailored to the structure of each problem in ways that yield faster convergence than general purpose optimization algorithms. Although certain applications motivate the algorithm development, we try to formulate solutions that benefit other applications too.
2) Theoretical analysis of the properties of these algorithms. We analyze convergence rate, spatial resolution, and statistical properties (especially noise) of these algorithms. Here the goal is often to find quantitative measures of image quality and algorithm performance.
3) Implementing the algorithms for clinical use and evaluating their performance with real data. One of the algorithms we have developed is currently in use at the UM Hospital for SPECT cardiac studies. Thousands of patients have been processed using this algorithm. Other algorithms are currently under investigation for possible routine clinical use.
4) Development of design criteria for imaging systems. The goal of this work is to address the question of how to choose imaging system design parameters to optimize image quality.
We perform the majority of this interdisciplinary work in collaboration with researchers in the UM Medical Center, in the EECS department, the Biomedical Engineering (BME) department, the Nuclear Engineering and Radiological Sciences (NERS) department, and in UM Biostatistics department. We also collaborate with researchers at other institutions and with industry.
The specific applications and problems that students investigate are evolving continuously at a pace too rapid to keep current on the web site. Applications of interest include MRI, X-ray CT, PET, SPECT, radiation therapy, and image registration. The list of submitted journal papers (and recent conference papers) under the publications link provide a glimpse of current work.
Please refer to the web site http://www.eecs.umich.edu/~fessler for more information.