Randoms simulation for dual head coincidence imaging

The high count rates in Dual Head Coincidence Imaging (DHCI) produce significant randoms fractions in the acquired data. The authors simulated random coincidence data from uniform cylinders in a DHCI system with 51 cm/spl times/37 cm detectors spaced 63 cm apart. The randoms were normalized as if they were true events and reconstructed using 3D reprojection. The simulated data are relatively independent of cylinder size and attenuation medium. The reconstructed random coincidences are broadly distributed throughout the field of view (FOV) but have some transaxial structure. They have a zero offset at the transaxial FOV edge, peak toward the center in the outer two-thirds of the FOV, and dip in the central third of the FOV. A 25% randoms fraction in a 20 cm diameter cylinder contributes an offset of less than 5% of the image intensity in a true+scatter image with 12 million counts and no attenuation correction. Attenuation correction without randoms subtraction produces significant errors in the reconstructed image. Attenuation correction applied to randoms-subtracted cylinder does not significantly enhance the image noise.