Comparison between segmented and nonsegmented attenuation correction on the HR+ tomograph

Segmentation of transmission (TR) images is implemented to improve signal noise in attenuation correction (AC) and allow shorter TR scans. This study compares (i) measurement of myocardial glucose metabolic rate (MRGlu), (ii) correction for myocardial wall thickness and (iii) measurement of lung density, by using AC derived from transmission images (a) unsegmented (US), (b) segmented (S) and (c) reconstructed with OSEM. All data were acquired on a CTI HR+ tomograph and processed with standard scanner software. Methods: TR scans were carried out on a multi-cylinder phantom containing substances of varying density and mean mu-values (cm
^-1
) were derived from TR images. A multi-compartment heart phantom was constructed as a simple simulation of varying myocardial wall thickness and left ventricular volume. MRGlu, using FDG (2D mode) and blood volume (C
¹⁵
O inhalation), was measured in 11 patients with coronary heart disease, following a 20 min TR scan. Results: Good linearity was found between NS mu-values and density but there were significant discrepancies with S over the typical range of lung density, and better agreement with OSEM. Overall mean ratios of MRGlu were within 2% for the three AC methods but there was systematic individual variation about the mean. The C.V. of regional MRGlu for each patient, using US AC, was not significantly different from S and OSEM. The heart phantom data demonstrated reasonable accuracy of correction for myocardial wall thickness. Conclusions: US AC on the HR+ gives more accurate measurement of the range of density encountered in the chest and more reliable measurement of MRGlu than S and OSEM. There appears to be no "statistical penalty" in the use of US AC, noise being dominated by that in emission data