Related Experiment Videos
Effect of smoothing during transmission processing on quantitative cardiac PET scans
N M Freedman1, S L Bacharach, R E Carson
1Department of Nuclear Medicine, National Institutes of Health, Bethesda, Maryland 20892, USA.
Unlabelled:
The effects of attenuation in cardiac PET are large and are produced by varied and inhomogeneous attenuating media. Although a measured attenuation correction can potentially provide an exact correction for attenuation, it introduces noise into the attenuation-corrected emission scan. Transmission smoothing reduces this noise but can introduce error of its own. This study investigates these errors in absolute and relative quantitation and estimates their magnitude in a clinical setting.
Methods:
Fluorodeoxyglucose cardiac PET scans of 24 subjects were processed using measured attenuation correction with different levels of transmission smoothing. Mean activity concentrations were determined in septal, anterior and lateral regions of the left ventricle at each level of transmission smoothing. A theoretical derivation of the effects of transmission smoothing is presented, so that the observed effects could be compared with theory-based predictions.
Results:
In addition to the reduction of noise, transmission smoothing produced two further effects: (a) a previously unreported reduction in noise-induced bias, which is beneficial and (b) introduction of errors due to bad estimates of attenuation correction factors resulting from smoothing over regions where attenuation changes. The first effect was observed over all regions of the left ventricle, whereas the second reduced counts primarily in the lateral wall. Twenty-millimeter smoothing reduced noise-induced bias by an average of 4% (compared with 6-mm smoothing). This same smoothing caused an additional 9% decrease in the lateral wall as a result of the adjacent lung-lateral wall boundary.
Conclusion:
Transmission smoothing reduces both noise and noise-induced bias, but near transitions between differently attenuating media (e.g., lung-myocardial borders) may produce errors in absolute and relative quantitation. The data presented here document the magnitudes of these effects, permitting one to ensure that artifactually introduced inhomogeneities are kept small.