Quantitative analysis of template-based attenuation compensation in 3D brain PET

An atlas-guided attenuation correction method was recently proposed for 3D brain positron emission tomography (PET) imaging eliminating the need for acquisition of a patient-specific measured transmission scan. The algorithm was validated through comparison to transmission-based attenuation correction (gold standard) using voxelwise statistical parametric mapping (SPM) analysis of clinical data. In contrast to brain 'activation' studies for which SPM is primarily developed, brain PET research studies often involve absolute quantification. In the preliminary validation study published earlier, there is no validation as to how such quantification can be affected by the two methods as the assessment was carried out by an SPM group analysis alone. It is quite important to demonstrate how the proposed method performs individually, particularly for diagnostic applications or individual quantification. In this study, we assess the quantitative accuracy of this method in clinical setting using automated volume of interest (VOI)-based analysis by means of the commercially available BRASS software. There is a very good correlation (R(2)=0.91) between the atlas-guided and measured transmission-guided attenuation correction techniques and the regression line agreed well with the line of identity (slope=0.96) for the grouped analysis of patient data. The mean relative difference between the two methods for all VOIs across the whole population is 2.3% whereas the maximum difference is less than 7%. No proof of statistically significant differences could be verified for all regions. These encouraging results provide further confidence in the adequacy of the proposed approach demonstrating its performance particularly for research studies or diagnostic applications involving quantification.