Predicting the transition from normal aging to Alzheimer's disease: A statistical mechanistic evaluation of FDG-PET data

The assessment of the degree of order of brain metabolism by means of a statistical mechanistic approach applied to FDG-PET, allowed us to characterize healthy subjects as well as patients with mild cognitive impairment and Alzheimer's Disease (AD). The intensity signals from 24 volumes of interest were submitted to principal component analysis (PCA) giving rise to a major first principal component whose eigenvalue was a reliable cumulative index of order. This index linearly decreased from 77 to 44% going from normal aging to AD patients with intermediate conditions between these values (r = 0.96, p < 0.001). Bootstrap analysis confirmed the statistical significance of the results. The progressive detachment of different brain regions from the first component was assessed, allowing for a purely data driven reconstruction of already known maximally affected areas. We demonstrated for the first time the reliability of a single global index of order in discriminating groups of cognitively impaired patients with different clinical outcome. The second relevant finding was the identification of clusters of regions relevant to AD pathology progressively separating from the first principal component through different stages of cognitive impairment, including patients cognitively impaired but not converted to AD. This paved the way to the quantitative assessment of the functional networking status in individual patients.