Dementia differentiation

We are applying advances in voxel-based multivariate analysis and machine learning to differentiate different forms of dementia, combining this with a unique reference set of data from subjects having Mild Cognitive Impairment (amyloid positive, amyloid negative), Alzheimer’s Disease (AD), Frontotemporal Dementia (FTD), Semantic Dementia, Corticobasal Syndrome, Lewy Body Disease, and other dementias, achieving excellent discrimination accuracies.  Advantages include recognition of relationships between affected brain regions, consolidation of multiple brain region changes into a single metric, and measures of reproducibility and predictive power.

Spatial patterns for dementia differentiation

Above is an illustration of differentiation between Normal, AD, and FTD conditions.  In the top images, blue regions are more hypometabolic in FTD than in AD or Normal, while in the bottom images, blue regions are more hypometabolic in AD than in FTD or Normal.  Individual subjects are scored quantitatively relative to the patterns that discriminate the states — the higher the score, the more the individual reflects the given pattern.  Red indicates relative preservation of metabolism.