Emission and Transmission Noise Propagation in Positron Emission Computed Tomography

摘要

Errors in positron emission computed tomograms are the result of noise propagated from three sources: (1) the statistical fluctuation in the positron coincidence events; (2) the statistical fluctuation in the incident transmission beam; and (3) the statistical fluctuation in the transmitted beam. The data for the transmission study in (2) and (3) are used to compensate for internal absorption of the distributed positron source. For the reconstruction of a circular phantom using the convolution algorithm, the percent root-mean-square uncertainty (%RMS) is related to the total measured positron events C and the incident photon flux per cm I sub 0 . Our derivation of the %RMS uncertainty based on the propagation of errors yields a simple expression: %RMS = sqrt K sub 1 /C + K sub 2 /I sub 0 . The constants K sub 1 = 4.52 x 10 exp 8 and K sub 2 = 1.48 x 10 exp 8 were determined for a 20 cm diameter disc based on computer simulation. The projection data were analytically calculated with an attenuation coefficient mu = 0.0958 cm exp -1 for 140 angles between 0 and pi . Poisson noise was added to the positron coincidence events, the incident transmission events I sub 0 , and the transmitted events. These results indicate that for a total number of incident transmission photons per cm of 2.0 x 10 exp 5 , the contrast resolution for a fixed spatial resolution is limited to 27% even with an infinite number of emission events. For a total of 10 exp 6 emission events the contrast resolution is 34%. (ERA citation 05:001839)