Segmented attenuation correction for PET

The authors describe a hybrid measured/calculated method for attenuation correction in positron emission tomography (PET). This unified reconstruction/segmentation method is based on a penalized weighted least-squares objective function that is minimized using iterative coordinate-descent. Two penalty functions are compared: one for a discrete object parameterization, the other for a continuous parameterization. Simulations demonstrate that the methods can reduce the additional emission image variance typically introduced by noisy attenuation correction factors.<<ETX>>