Dead-time correction for a rotating rod normalization in a cylindrical PET system

Depending on the geometry between the rotating coincidence rod source, the PET detectors and the rod activity, a variable amount of block dead-time is found in a PET system during normalization. This dead-time is driven by the relative location between the rod source and the crystals in the block detector. Normalization scans were acquired on a GE Discovery ST PET-CT system with 3 rod activities of duration such that the total acquired counts (T+S+R) were held constant. To develop a model of the dead-time, acquisitions at six static source locations, centered over each crystal in a single block detector, were acquired for each of the rod activity levels. The resultant block busy data were analyzed such that the profile of block busy as the rod traversed all lines-of-response (with respect to the said block) was found. The profile was fit with a Gaussian function and parameterized by the FWHM and amplitude of the Gaussian. For image analysis, a 20 cm uniform cylinder and a whole-body patient scan were analyzed in reconstructed image space. The datasets were reconstructed with each normalization correction, then with the normalization corrected for dead-time. A model has been found and is described allowing application of dead-time correction to a rotating rod normalization scan based on measured block-busy in the normalization raw data. The model corrects for dead-time effects in the normalization. Analysis of image quality impact shows that bias and variance effects are reduced.