Regularization methods in iterative algorithms for variance reduction on compressed sinogram random coincidences

The Poisson model based image reconstruction algorithms in PET require the estimation of the mean value of random coincidences. The random data are often acquired with a delayed coincidence technique, and expected randoms are estimated through variance reduction (VR) of measured delayed coincidences. In the past we have developed iterative VR algorithms that estimate singles rates from random compressed sinograms. The final estimation of singles rates can be noisy in short scans. In the present work, we apply commonly used and specific regularization methods. A sequential coordinate ascent algorithm is used to maximize the penalized Poisson Likelihood objective function. Measured data from a Siemens TruePoint clinical scanner are used to validate the algorithm performance.