AUTHOR=Huang Sheng-Yao , Hsu Jung-Lung , Lin Kun-Ju , Hsiao Ing-Tsung 

TITLE=A Novel Individual Metabolic Brain Network for 18F-FDG PET Imaging

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 14 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2020.00344

DOI=10.3389/fnins.2020.00344

ISSN=1662-453X

ABSTRACT=Metabolic brain network analysis based on graph theory using FDG PET imaging is potentially useful for investigating brain activity alternation due to metabolism changes in different stages of Alzheimer's disease (AD).  Most studies in constructing metabolic network were based on group-data.  Here to investigate individual metabolic network abnormalities, a novel approach in building an individual metabolic network was proposed.
Method First, a weighting matrix was calculated based on the interregional effect size difference of mean uptake between a single subject and an average normal controls (NC).  Then the weighting matrix for a single subject was multiplied by a group-based connectivity matrix from a NC cohort.  To study the performance of the proposed individual metabolic network, inter- and intra-hemispheric connectivity patterns in the groups of NC, sMCI (stable mild cognitive impairment), pMCI (progressive mild cognitive impairment) and AD using the proposed individual metabolic network were constructed and compared with those from the group-based results.  The network parameters of global efficiency, clustering coefficient, lambda and gamma, and the network density score (NDS) in the default-mode network (DMN) of generated individual metabolic network were estimated and compared among the disease groups in AD.
Result Our results show that the intra and inter-hemispheric connectivity patterns estimated from our individual metabolic network are similar to those from the group-based method.  In particular, the key patterns of occipital-parietal  inter-regional connectivity deficits detected in the groupwise network study for differentiating different disease groups in AD were also found in the individual network.  A reduction trend was observed for network parameters of global efficiency and clustering coefficient, and also for the NDS from NC, sMCI, pMCI and AD.  There was no significant difference between NC and sMCI for all network parameters.
Conclusion 
We proposed a novel method in constructing the individual metabolic network using a single-subject FDG PET image and a group-based normal control connectivity matrix. The result has shown the effectiveness and feasibility of the proposed individual metabolic network in differentiating different disease groups in AD.  Future study should include investigation of inter-individual variability, and the correlation of individual network features to disease severities and clinical performance.