Detection of dementia-related hypometabolism using two different age-adjusted reference FDG- PET databases

Objectives Quantitation of reduced FDG uptake in a patient with dementia is often performed by single subject analysis compared to an FDG reference database. Since FDG uptake in human brain changes with age, ideally, age-matched subjects should be used as controls. However in practice, the reference database consists of normal subjects covering a wide range of age. In this case, the age effect can be modelled by regression leading to an age-adjusted reference FDG database as implemented in the Alzheimer's Discrimination Tool of PMOD (PALZ) (1). This approach was used in our previously constructed brain FDG database (NC37) (2) in order to construct an age-adjusted database (NC37a) for comparison to the PALZ reference database.

Methods A parametric map of r-values (r-map) was calculated in SPM8 by regression analysis of the NC37 with age. Age-adjustment of the NC37 database was performed by correcting each image of the 37 subjects using r-map. Twenty FDG datasets of patients (age range: 49 - 84 y) with Alzheimer’s disease (AD) were tested using either PALZ or the SPM8 single subject analysis and NC37a. Each maximum t-value and the PETscore (1) were reported.

Results Maximum t-values of 20 AD datasets using PALZ-like analysis with NC37a were higher (mean: 8.13, range: 5.86 - 11.12) than using PALZ (mean: 7.03, range: 4.56 - 9.44). PETscore was slightly higher with PALZ (mean: 2.66, range: 1.27 - 3.58) compared to NC37a (mean: 2.61, range: 1.42 - 3.72), in particular in younger AD patients (age < 60 y).

Conclusions Despite different constitution of the PALZ database and the NC37a database, the results of the analysis of AD-related hypometabolism in 20 AD datasets were similar. Therefore, diagnosis of dementia using our age-adjusted database NC37a is feasible. Age-dependence is more prominent in PALZ as compared to NC37a leading to higher PETscores (average: 8 %) in younger AD patients. Higher t-values with NC37a as compared to PALZ may be explained by a smaller standard deviation in our reference database.