Addressing the third gamma problem in PET

PET brings the promise of quantitative imaging of the in-vivo distribution of any positron emitting nuclide, a list with hundreds of candidates. All but a few of these, the "pure positron" emitters, have isotropic coincident gamma rays that give rise to misrepresented events in the sinogram and in the resulting reconstructed image. Of particular interest are /sup 10/C, /sup 14/O, /sup 38/K, /sup 52m/Mn, /sup 60/Cu, /sup 61/Cu, /sup 94m/Tc, and /sup 124/I, each having high-energy gammas that are Compton-scattered down into the 511 keV window. The problems arising from the "third gamma," and its accommodation by standard scatter correction algorithms, were studied empirically, employing three scanner models (CTI 933/04, CTI HR+ and GE Advance), imaging three phantoms (line source, NEMA scatter and contrast/detail), with /sup 18/F or /sup 38/K and /sup 72/As mimicking /sup 14/O and /sup 10/C, respectively, in 2-D and 3-D modes. Five findings emerge directly from the image analysis. The third gamma: 1) does, obviously, tax the single event rate of the PET scanners, particularly in the absence of septa, from activity outside of the axial field of view; 2) does, therefore, tax the random rate, which is second order in singles, although the gamma is a prompt coincidence partner; 3) does enter the sinogram as an additional flat background, like randoms, but unlike scatter; 4) is not seriously misrepresented by the scatter algorithm which fits the correction to the wings of the sinogram; and 5) does introduce additional statistical noise from the subsequent subtraction, but does not seriously compromise the detectability of lesions as seen in the contrast/detail phantom. As a safeguard against the loss of accuracy in image quantitation, fiducial sources of known activity are included in the field of view alongside of the subject. With this precaution, a much wider selection of imaging agents can enjoy the advantages of positron emission tomography.