A set of prototype biomarkers to capture prodromal Alzheimer's disease pathology: Neurophysiological graph metrics
Effective treatments of Alzheimer's disease depend on sensitive and reliable methods to detect early changes in AD pathology and response to interventions. Earlyy stages of AD pathogenesis are charactterised with synaptic dysfunction, imparied information processing before cognitive decline and atrophy are observed.
E/MEG is a direct and temporally precise measure of neuronal activity. In this analysis we use E/MEG to capture early signs of synaptic dysfunction and provide neurophysiological biomarkers of AD which could aid future clinical trials.
Deep and Frequent Phenotyping Feasability Study is a multicentre pilot study that gathered frequent multimodality measurements from participants who had positive AD biomarkers as measured by CSF analyis. The multimodality measurements included MR, MEG, PET, CSF, gait, neuropsychological scores and blood sampling.
In this analysis I focus on the MEG (resting state eyes open) and PET (AV1451) data, and use graph theory as a way to summarise connectivity changes that occur in different frequency bands in early AD.
Cartoon depicting early AD pathogenesis which is characterised by the spread of amyloid beta plaques around synapses and hyperphosphorylated tau spread and accumulation in the cell. E/MEG could be more sensitive than other modalities to pick up the early synaptic dysfunction.
The group mean for AV1451 BPND. The distribution of binding is characteristic of AD lighting up in bilateral posterior, inferior temporal, parietal areas, as well as putamen and precuneus.
We computed three graph metrics that showed the minimal intercorrelations across 5 frequency bands. Strength, as the name suggests captures the strength of connectivity each node has with the remaining nodes. Closeness centrality measures the functional distance among the nodes, and is a measure of local efficiency. Lastly, participation coefficient captures the diversity of intermodular connections of each node.
We extracted the mean AV1451 BPND values for each node, and related the means to the graph metrics both at the whole brain and at the lobar levels.
At the whole brain level, we found that as tau accumulates,
1) participation of the nodes in multiple modules is reduced in the theta band;
2) connectivity strength is reduced in the alpha band;
3) local efficiency of information transfer is reduced in the gamma band.
At the lobar level:
1) participation coefficient and strength effects were driven by all nodes, albeit strongest is the temporal nodes;
2) closeness centrality effect was significantly driven by the frontal and limbic nodes.
Scatterplots at the top show the whole brain level results for the band that show significant results. The plots at the bottom split the scatterplots by the functional lobes, for the effects above.
Results show that, even with a small sample size of 12 people, E/MEG is able to capture network connectivity biomarkers of early AD, which significantly overlap with tau accumulation across the cortex.