Quantitative Ga-67 SPECT imaging

Tumor avidity for Ga-67 is known to be correlated with the histopathologic stage of disease, yet quantitative imaging has remained elusive because Ga-67 images are seriously degraded by several factors. We have previously optimized and evaluated on projection images a general spectral (GS) approach to compensate for patient scatter and Pb x-rays from the collimator. We also recently developed a joint, multienergy iterative algorithm (OSEM-J) to compensate for nonuniform attenuation and for distance- and energy-dependent collimator resolution. In the work reported here, we evaluated the overall activity quantitation performance achieved by combining GS with OSEM-J, and compared this to the (almost ideal) performance obtained when reconstructing primary unscattered photons, as well as to that obtained with another proposed iterative approach (OSEM-S) in which the sum of the two lowest photopeak windows was reconstructed using a single, energy-weighted attenuation map. We simulated 96 projections of an anthropomorphic phantom containing seven 2-cm-diameter tumors with Ga-67 concentration values 4 to 8 times that of the local background. Sixteen noise realizations containing lesions and 16 without lesions were generated, and least-squares estimates of tumor activity were calculated for each reconstruction method. The mean bias of tumor estimates for primary+OSEM-J, GS+OSEM-J, and GS+OSEM-S was, respectively, 1.2%, 1.6%, and 8.0% The average precision of the corresponding estimates was 15.0%, 19.1%, and 24.8% These results suggest that use of a spectral-based estimate of contaminant photons within a joint multiple-energy iterative reconstruction leads to nearly optimal performance in quantitation of Ga-67 tumor activity.