Purpose: [(18)F]fluorodeoxyglucose (FDG) PET imaging of the brain can be used to assist in the differential diagnosis of dementia. Group differences in glucose uptake between patients with dementia and controls are well-known. However, a multivariate analysis technique called scaled subprofile model, principal component analysis (SSM/PCA) aiming at identifying diagnostic neural networks in diseases, have been applied less frequently. We validated an Alzheimer's Disease-related (AD) glucose metabolic brain pattern using the SSM/PCA analysis and applied it prospectively in an independent confirmation cohort.
Methods: We used FDG-PET scans of 18 healthy controls and 15 AD patients (identification cohort) to identify an AD-related glucose metabolic covariance pattern. In the confirmation cohort (n=15), we investigated the ability to discriminate between probable AD and non-probable AD (possible AD, mild cognitive impairment (MCI) or subjective complaints).
Results: The AD-related metabolic covariance pattern was characterized by relatively decreased metabolism in the temporoparietal regions and relatively increased metabolism in the subcortical white matter, cerebellum and sensorimotor cortex. Receiver-operating characteristic (ROC) curves showed at a cut-off value of z=1.23, a sensitivity of 93% and a specificity of 94% for correct AD classification. In the confirmation cohort, subjects with clinically probable AD diagnosis showed a high expression of the AD-related pattern whereas in subjects with a non-probable AD diagnosis a low expression was found.
Conclusion: The Alzheimer's disease-related cerebral glucose metabolic covariance pattern identified by SSM/PCA analysis was highly sensitive and specific for Alzheimer's disease. This method is expected to be helpful in the early diagnosis of Alzheimer's disease in clinical practice.