Metabolic heterogeneity in ADNI PET subjects

Objectives The Alzheimer’s Disease Neuroimaging Initiative (ADNI) is a multi-center effort to establish biomarkers for Alzheimer’s Disease (AD). Our study is to investigate metabolic heterogeneity among ADNI subjects who were recruited based on clinical diagnosis.

Methods Baseline FDG-PET scans of ADNI subjects with AD (n=97,age=76±7y, MMSE=24±2) and normals (NC;n=97, age=76±5y, MMSE=29±1) were assessed. Two classification systems were employed for consensus characterization of PET: 1) 3D-SSP database-lookup approach with image probability matching (JNM 1996) and 2) stereotactic voxel-based classification with partial least squares, resulting in observer-independent classifications of NC, AD, frontotemporal dementia FTD, and Lewy body dementia LBD. Discordant cases between ADNI and classification were further evaluated by 3D-SSP Z-score maps.

Results Discordant cases between ADNI diagnosis and FDG PET classification were found in 23 AD cases (24%) and 6 controls (6%). Among discordant AD cases, 12 were labelled as normal and 11 as FTD. Among discordant NC cases, 3 showed AD and FTD pattern, respectively. 3D-SSP Z-maps of the discordant cases showed atypical or fronto-temporal dominant findings in AD and confirmed ADNI diagnosis in 6 of the 23 cases. In 4 controls, AD or FTD pattern was found. One expert system also identified LBD in 5 cases.

Conclusions This study indicates heterogeneous metabolic patterns among clinically diagnosed subjects in ADNI data. This heterogeneity may indicate either inaccurate clinical diagnosis or presence of atypical metabolic patterns. Precaution needs to be taken when investigating on ADNI data.