Contribution of scatter and attenuation compensation to SPECT images of nonuniformly distributed brain activities.

UNLABELLED: Correction of scatter and attenuation is essential for quantitative SPECT. In this work, we evaluated the accuracy gained from a method of transmission-dependent convolution subtraction (TDCS) in the quantitation of activity that is highly concentrated in the striatum (STR). METHODS: SPECT data were acquired from an (123)I-containing phantom with a constant activity in the STR but differing background (BKG) activities, so as to simulate various STR/BKG ratios (19.7:1, 9.7:1, 4.8:1, 1.9:1, and 1:1). In a study of healthy humans (n = 6), a transmission scan followed by an emission scan was performed 24 h after injection of (123)I-2beta-carbomethoxy-3beta-(4-iodophenyl)-tropane ((123)I-beta-CIT). All SPECT data was reconstructed with ordered-subset expectation maximization. TDCS was applied for scatter correction. Values of activity in the STR and occipital lobe (for BKG) were used to calculate binding potential V(3)" (= [STR - BKG]/BKG). The effect of SPECT collimator dependency on scatter correction was also evaluated for 6 collimators from 3 different SPECT cameras in the phantom experiment. RESULTS: Scatter correction in the phantom experiment increased the measured values of STR activity (36.2%), resulting in a substantial increase in V(3)" (66.1%). Scatter and attenuation corrections with recovery correction showed an overall bias of -7.3% for the STR, -4.0% for BKG activity, and -7.8% for V(3)". TDCS corrections of phantom activities were relatively uniform for the 6 different collimators, with variabilities of <5.5% for the STR and <3.0% for BKG activities. TDCS correction of human (123)I-beta-CIT images was of a similar, although slightly larger, magnitude than for the phantom data, with increased V(3)" values of 9.4 +/- 2.3 and 4.9 +/- 0.6, with and without scatter correction, respectively. CONCLUSION: The TDSC method significantly improved the accuracy of SPECT images with a nonuniform distribution of activity highly concentrated in central regions. The value of V(3)" was significantly increased in phantom and human data, with most of the improvement derived from an increase in STR activity. This scatter correction method was approximately equally useful with data from the 6 different collimators and is recommended for more accurate quantitation of nonuniformly distributed brain activities.