Test-retest reproducibility of structural and proxy estimates of brain connectivity at rest

While structural connectivity (SC) reflects actual neural connections, proxy estimates of brain connectivity such as functional connectivity (FC) from functional MRI, intersubject covariance of regional gray matter volume (GMV_cov) from structural MRI, and intersubject covariance of regional 18F-fluorodeoxyglucose uptake (FDG_cov) from PET are derived from statistical dependencies between regional measurements. Understanding reproducibility of these estimates is crucial for capturing physiological and pathological changes in brain connectivity. Here, we compared test-retest reproducibility of group-level SC, FC, GMV_cov, and FDG_cov in the same 55 healthy subjects at rest using a simultaneous PET/MRI acquisition protocol. Sparsity-based thresholding resulted in coefficients of variation of 2.7 %, 3.1 %, 3.6 %, and 5.1 % for SC, FDG_cov, GMV_cov, and FC, respectively. Strength-based thresholding resulted in relative proportion of repeatedly (test-retest) present connections (PRPC) of 77 %, 56 %, and 54 % for GMV_cov, FDG_cov, and FC, respectively. Absolute PRPC were 2.50 %, 0.64 %, and 0.25 % for FDG_cov, FC, and GMV_cov, respectively. Overall, stronger connections exhibited higher reproducibility, arguing in favour of thresholding as a necessary analytical step. SC was most reproducible. Among the proxy estimates, the highest absolute PRPC was found for FDG_cov. FDG_cov enables to study brain connectivity in a reproducible manner over the largest part of the brain.