Inter-subject FDG PET Brain Networks Exhibit Multi-scale Community Structure with Different Normalization Techniques

• Published in: Annals of biomedical engineering Vol. 46; no. 7; pp. 1001 - 1012
• Main Authors: Sperry, Megan M., Kartha, Sonia, Granquist, Eric J., Winkelstein, Beth A.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.07.2018; Springer Nature B.V
• Subjects: Animals; Biochemistry; Biological and Medical Physics; Biomedical and Life Sciences; Biomedical Engineering and Bioengineering; Biomedicine; Biophysics; Brain; Cerebellum; Cerebellum - diagnostic imaging; Cerebellum - physiology; Classical Mechanics; Clustering; Communities; Community structure; Correlation; Fluorine isotopes; Glucose; Glucose-6-Phosphate - analogs & derivatives; Glucose-6-Phosphate - pharmacology; Group size; Image acquisition; Imaging techniques; Mesoscale phenomena; Models, Neurological; Modularity; Multiscale analysis; Nerve Net - diagnostic imaging; Nerve Net - physiology; Networks; Neural networks; Neuroimaging; Positron emission; Positron emission tomography; Rats; Size effects; Stability analysis; Tomography; Visual cortex; Visual Cortex - diagnostic imaging; Visual Cortex - physiology; Networks; Brain; Modularity; FDG PET
• ISSN: 0090-6964; 1573-9686; 1573-9686
• DOI: 10.1007/s10439-018-2022-x

Inter-subject networks are used to model correlations between brain regions and are particularly useful for metabolic imaging techniques, like 18F-2-deoxy-2-(18F)fluoro- d -glucose (FDG) positron emission tomography (PET). Since FDG PET typically produces a single image, correlations cannot be calculated over time. Little focus has been placed on the basic properties of inter-subject networks and if they are affected by group size and image normalization. FDG PET images were acquired from rats ( n  = 18), normalized by whole brain, visual cortex, or cerebellar FDG uptake, and used to construct correlation matrices. Group size effects on network stability were investigated by systematically adding rats and evaluating local network connectivity (node strength and clustering coefficient). Modularity and community structure were also evaluated in the differently normalized networks to assess meso-scale network relationships. Local network properties are stable regardless of normalization region for groups of at least 10. Whole brain-normalized networks are more modular than visual cortex- or cerebellum-normalized network ( p  < 0.00001); however, community structure is similar at network resolutions where modularity differs most between brain and randomized networks. Hierarchical analysis reveals consistent modules at different scales and clustering of spatially-proximate brain regions. Findings suggest inter-subject FDG PET networks are stable for reasonable group sizes and exhibit multi-scale modularity.