 |
Patrick La Rivière |
Patrick La Rivière is pursuing a Ph.D. in Medical Physics at the University of Chicago. A native of Montreal, Canada, Patrick received an A.B. magna cum laude in Physics from Harvard University in 1994, where he also served as editor-in-chief of the weekly campus newspaper, the Harvard Independent. In his senior year, Patrick was awarded the Lionel de Jersey-Harvard scholarship to Cambridge University, which allowed him to live in John Harvard’s own rooms at Emmanuel College. While at Cambridge, he worked with Prof. Michael Redhead in the History and Philosophy of Science department on questions relating to the conceptual foundations of quantum mechanics and specifically on analyzing and developing relativistic formulations of the Einstein- Podolsky-Rosen paradox.
|
Patrick La Rivière |
Upon his return to North America in the Fall of 1995, Patrick worked as a technical writer at the Chicago Academy of Sciences, a science museum with a focus on the natural history of the Midwest and a strong commitment to science outreach education. In January, 1996, he enrolled in the Graduate Program in Medical Physics in the Department of Radiology at the University of Chicago, where he works under Assistant Prof. Xiaochuan Pan.
His general area of research interest is in tomographic reconstruction in nuclear medicine and computed tomography. Specifically, he is working on sparse-data problems that require smoothing and interpolation for adequate reconstruction, such as few-view emission computed tomography and helical computed tomography (CT). This work has also prompted him to investigate fundamental properties of a number of smoothing and interpolation approaches including Fourier-based techniques, cubic-spline techniques, as well as more familiar linear interpolation approaches.
The development of algorithms for few-view emission tomography reconstruction has been carried out with an eye toward a particular application: dedicated single-photon emission computed tomography (SPECT) scintimammography (SMM). SMM is a nuclear medicine technique involving the injection of a radiotracer (usually Tc-99m-sestamibi) that accumulates preferentially in malignant tissue. It can thus be used to differentiate benign from malignant abnormalities found by x-ray mammography as well as to provide outright detection of malignancies in women with radiographically dense breasts. While most SMM is performed with a planar imaging geometry, Patrick has shown that a dedicated SPECT geometry, in which a small gamma camera revolves around the breast, is likely to provide improved lesion detectability. The ability to reconstruct dedicated SPECT SMM images from a smaller number of views than is usually used could help reduce imaging time, thereby reducing motion artifacts, increasing patient throughput, and generally making the technique more viable as a high-volume follow-up test to mammography. This project has been suported by a Department of Defense Breast Cancer Research Program predoctoral award.
For this few-view tomography problem, Patrick has pursued a strategy involving sinogram preprocessing, in which each projection is first smoothed using spline-based techniques and then additional projections are interpolated, again using spline-based techniques, prior to reconstruction by FBP. The spline-based projection smoothing technique is a novel application of roughness-penalized nonparametric regression using an explicit Poisson model. Patrick has shown that this approach induces a resolution-noise tradeoff that is superior to that obtained with the use of Fourier-domain apodization windows and also superior to that obtained by use of a roughness-penalized weighted least squares objective function. He has also shown that through judicious choice of the model’s so-called link function, the technique can yield images with essentially uniform, isotropic resolution, whereas most statistically based smoothing and reconstruction approaches lead to nonuniform, anisotropic resolution. This smoothing approach is generally applicable to noisy tomographic data, and is not in any way limited to few-view situations.
The helical CT strand of the work involves the development of Fourier-based approaches to the longitudinal interpolation step that is necessary to convert the helical projection data into a form that can be reconstructed by use of transverse image reconstruction algorithms. Specifically, two new and related approaches have been developed that under certain conditions allow for the realization of the long-stated goal of achieving essentially isotropic resolution and noise properties in reconstructed helical CT volumes. Both approaches exploit the fast Fourier transform and the Fourier shift theorem to generate from the helical projection data a set of fan-beam sinograms corresponding to equispaced transverse slices. Slice-by-slice reconstruction is then performed by use of two-dimensional fan-beam algorithms. The first approach, called 360FT, makes use only of the directly measured projection data, but an extension called 180FT exploits the redundancy of fan-beam data acquired over 3600 to generate a second set of longitudinal samples at each projection angle and bin. For low to moderate pitches, these approaches, and particularly the 180FT approach, have been shown to produce reconstructed volumes with much more istropic resolution and aliasing properties than do existing approaches based on the use of linear interpolation. The approaches have also been shown to have more favorable noise uniformity properties than do currently used approaches.
Patrick anticipates graduating in the year 2000 and pursuing an academic career involving research in basic medical imaging physics and engineering.
Patrick La Rivière can be reached at the Department of Radiology, 5841 S. Maryland Ave., MC-1037, The University of Chicago, Chicago, IL 60637; Phone: (773) 702-6975; Fax: (773) 702-5986; E-mail: pjlarivi@midway.uchicago.edu.