Positron Emission Tomography
Research Summary
Our work in PET is focused on optimizing image quality through improved statistical and physical modeling of data acquisition and system response. We have addressed this problem using MAP (maximum a posteriori) estimation combining detector response modeling with accurate statistical modeling of random and scatter correction with rapidly convergent algorithms. Using spatially variant priors we are able to achieve count-independent image resolution that can be controlled by a single global smoothing parameter. By modeling the positron emissions from each voxel as a time inhomogeneous Poisson process, we extend the MAP approach to 4D (3 spatial + 1 temporal) reconstruction.
Our current emphasis is on extending the MAP approach to time of flight (TOF) PET data and exploring the trade-off between resolution/noise performance of rebinning fully 3D TOF data to either 2D or 3D non-TOF data prior to MAP reconstruction. We are collaborating in this effort with Frank DiFilippo at the Cleveland Clinic.
Our other current area of interest is in using dynamic data to better characterize tracer uptake in tumors. In collaboration with Quanzheng Li (MGH, and a former lab member) we are developing techniques for whole body Patlak imaging using novel methods for Patlak parameter estimation from partial data to enable parametric imaging from data using multiple bed positions.
Current Collaborators
Peter Conti, Director of Clinical PET, Professor Radiology.
Quanzheng Li, Assistant Professor, Dept. Radiology. MGH/Harvard.
Frank DiFilippo, PhD, Dept. Nuclear Medicine and Biomedical Engineering, Cleveland Clinic
Research Support
5R01EB010197: OPTIMIZED IMAGE RECONSTRUCTION FOR TIME-OF-FLIGHT PET, PI: Richard Leahy
1R01EB013293: QUANTITATIVE METHODS FOR CLINICAL WHOLE BODY DYNAMIC PET, PI: Quanzheng Li (MGH)
5R21CA149587: AN INTEGRATED STATISTICAL FRAMEWORK FOR LESION DETECTION USING DYNAMIC PET , PI: Quanzheng Li (MGH)