Posterior cortical atrophy phenotypic heterogeneity revealed by decoding ¹⁸F-FDG-PET.

Area of Science:

• Neuroscience
• Radiology
• Machine Learning

Background:

• Posterior cortical atrophy (PCA) is a neurodegenerative syndrome with diverse clinical presentations due to variable involvement of visual and cognitive systems.
• ¹⁸F-fluorodeoxyglucose (FDG)-PET is a sensitive imaging biomarker for detecting regional brain damage and capturing neurodegenerative patterns at the individual level.

Purpose of the Study:

• To investigate the heterogeneity of posterior cortical atrophy by analyzing inter-individual differences in FDG-PET imaging.
• To associate distinct neuroimaging patterns with clinical, neuropsychological, and biomarker data in PCA patients.

Main Methods:

• Analyzed FDG-PET data from 91 posterior cortical atrophy participants.
• Applied unsupervised machine learning (eigen-decomposition) to identify principal axes of inter-individual variation in brain glucose metabolism ('eigenbrains').
• Correlated 'eigenbrain' patterns with demographic, clinical, neuropsychological, and Alzheimer's disease biomarker data; utilized NeuroSynth for functional characterization.

Main Results:

• Eight 'eigenbrains' explained over 50% of the inter-individual variability in FDG-PET uptake, with left and right hemispheric patterns accounting for 24% of the variance.
• Specific 'eigenbrain' patterns correlated with distinct clinical features, including aphasia/apraxia (left hemisphere) and visual agnosias (right hemisphere).
• Age of onset was associated with different 'eigenbrain' patterns, such as limbic-predominant versus frontoparietal.

Conclusions:

• Machine learning analysis of FDG-PET reveals distinct neurobiological subtypes within posterior cortical atrophy, correlating with specific clinical and cognitive profiles.
• Focusing on inter-individual differences in neuroimaging is crucial for understanding the heterogeneity of neurodegenerative syndromes like PCA.
• These findings highlight the potential of 'eigenbrain' analysis to capture biologically relevant variations in neurodegeneration.