Abstract
1296
Objectives Statistical z-mapping such as 3D-SSP is widely used to interpret and diagnose brain FDG PET images of patients with suspected neurodegenerative or psychiatric disorders, in which subject images are compared pixel-wise with normal data base after spatial normalization. While such software is usually equipped with its standard normal data base, aging effect, if any, may under- or overestimate the offset value and lead to misdiagnosis of aged subjects. The effect of age on the normal FDG distribution and its effect on the z-map were investigated.
Methods A total of 50 normal subjects with the age ranging from 45 to 77 underwent brain FDG PET in 5 PET centers using 5 different scanners. The images were spatially normalized with 3D-SSP and those that exhibited normalization error or evident abnormal distribution were removed as outliers. The subjects were classified into 5 age groups, and the images were examined for age effect. Then, z-map was created for each subject using the normal data base consisting of all the 177 subjects of this study as well as using one of the age groups as the normal data base.
Results After correction for site difference, the mean image for each age group demonstrated age-related decrease in the frontal lobe, reduction by 28.6 % in 70s as compared with 50-60s, while the standard deviation was comparable between age groups. When the normal data base of the entire subjects were used, 17 % of the subjects of 70s presented a decrease of z-mean>=2.5 in the frontal lobe as quantified with SEE program (Ann Nucl Med 17:4,2003), while it was 2.7 % for the subjects of 60s. An identical subject of 70s presented false-positive hypometabolism when the normal data base was made of subjects with 50s, but was interpreted as normal when the database was made of 70s.
Conclusions For the diagnosis of frontal lobe hypometabolism using 3D-SSP, age-matched normal data base may be used to avoid misinterpretation