Randoms Correction Based on Time Sampled Compressed Delayed Events

One of the techniques used for the estimation of random contributions to prompt data is delay window acquisition. Measured delayed data contain high-level noise and variance reduction methods are exploited to generate mean random data. These data are used in the image reconstruction algorithms to model the contribution of background events. The classical equation where the randoms rate is a product of two singles rates is the foundation for delays variance reduction methods: histogrammed delays are decomposed into singles rates. Nevertheless, the singles rate may change over time, as in the case of Continuous Bed Acquisition (CBM). Currently, CBM delays data are acquired over the duration of the whole scan. An approximate method is used to estimate mean randoms. While time varying singles are straightforwardly accommodated in the random model, their estimation requires acquisition of a time dependent delays sinogram. This would be resource consuming due to the large sinogram size. Fortunately, in singles rate estimations, only a compressed form of delays is necessary to form the cone-sum, and this is computable directly from the list mode file.An axially long field of view (ALFOV) scanner, the Biograph Vision Quadra (Siemens Healthineers), was used in the investigation along with a Quality Control (QC) phantom. A relatively axially short cylinder phantom traveled through the scanner FOV, resulting in significantly varying detector delays rates. The use of time sampled delay cone-sums resulted in mean random sinogram generation, which is in quantitative agreement with the measured delays data.