Spatiotemporal Reconstruction of PET Data

• Activity Number: 298
• Type: Invited
• Date/Time: Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Graphics*
• Author(s): Richard Leahy*+ and Thomas Nichols
• Affiliation(s): University of Southern California and University of Michigan
• Keywords: Medical imaging ; Image reconstruction ; inverse problems ; Poisson Process ; Splines ; Positron Emission Tomography

Abstract

Positron Emission Tomography (PET) is a method for imaging the function of living tissue. Raw PET data consists of projections through the object which must be "backprojected" or reconstructed to form images. In 1985, Shepp & Vardi introduced the use of EM for reconstruction of PET images from the raw projection data. Since then many advances have been made which have resulted in improved spatial resolution. Temporal resolution, however, has been limited by the hardware, which integrates the data over five-minute or longer intervals.

We us data from new machines which produce list mode data, raw data with millisecond time resolution. We use list mode data to estimate spatiotemporal "movies" of tracer intensity. We model the data with an inhomogeneous Poisson process, where the rate function is parameterized with a B-spline. We use penalized likelihood, with penalty terms that regularize spatial and temporal smoothness and enforce positivity in the rate functions. We characterize the method with simulations and demonstrate it with real data.