An attenuation measurement technique for rotating planar detector positron tomographs.

This paper presents a new attenuation measurement technique suitable for rotating planar detector positron tomographs. Transmission measurements are made using two unshielded positron-emitting line sources, one attached to the front face of each detector. Many of the scattered and accidental coincidences are rejected by including only those coincidences that form a vector passing within a predetermined distance of either line source. Some scattered and accidental coincidences are still included, which reduces the measured linear attenuation: in principle their contribution can be accurately estimated and subtracted, but in practice, when limited statistics are available (as is the case with the multi-wire Birmingham positron camera), this background subtraction unacceptably increases the noise. Instead an attenuation image having the correct features can be reconstructed from the measured projections. For objects containing only a few discrete linear attenuation coefficients, segmentation of this attenuation image reduces noise and allows the correct linear attenuation coefficients to be restored by renormalization. Reprojection through the segmented image may then provide quantitatively correct attenuation correction factors of sufficient statistical quality to correct for attenuation in PET emission images.