The goal of the proposed research is to increase the accuracy of the identification of patients with coronary artery disease by single photon emission computed tomographic (SPECT) cardiac perfusion imaging with Tc-99m sestamibi. This will be accomplished by correcting for a number of the degradations inherent in SPECT imaging. The corrections will be separated to handle each degradation efficiently and accurately. First, the nonuniform attenuation of Tc-99m's photons in the chest will be corrected by using patient specific attenuation maps with iterative correction of the slices. The attenuation maps will be obtained either from transmission images acquired simultaneously with the emission images using a line source opposed to a fan beam collimator on one head of a three-headed SPECT system, or the segmentation of body and lung regions in scatter and photopeak window images. Second, scatter will be corrected by subtracting an estimate of the scatter obtained by the dual photopeak window method from the initial emission projections. Third, the effects of distance dependent collimator blurring will be redressed using frequency distance principal (FDP) restoration filtering. The FDP allows for the unique determination of th blurring function, and hence its correction, in the Fourier transform of the sinograms of the scatter corrected slices. Use of restoration filters which favor improvement in spatial resolution over noise suppression will be investigated. The residual, approximately stationary and isotropic blurring will be included in the projector/backprojector pair. Fourth, ECG gating will be used to """"""""freeze"""""""" cardiac motion, and allow assessment of function. Finally, iteratively reweighted least squares, variable conductance diffusion, or Gibbs priors will be used with iterative correction of attenuation to perform a nonlinear suppression of noise which maintains edges in the reconstructed slices. Through use of pre-correction of the projections for scatter and distance dependent collimator blur, the computational load of iterative correction will be greatly reduced. The influence of the pre-corrections on the Poisson noise of the projection will be investigated quantitatively and in ROC studies. The impact of these corrections on diagnostic accuracy will be determined through ROC studies in Monte Carlo simulations of SPECT imaging of the chest using human observers, and sensitivity and specificity measured in comparison to the results from cardiac catheterization in patients.
Showing the most recent 10 out of 27 publications
