Any large genome-wide association study corrects for population structure, in addition to restricting the sample to subjects of caucasian ethnicity. This is for the most part a practical necessity since mankind’s migration pattern across the globe has left its marks on our genome. Not correcting for these effects would lead to false-positive genetic associations. However, the topic of correcting for these population structure effects has for the most part not reached the field of neuroimaging.
It has been shown in earlier work that head and brain morphology is markedly different between continental genetic ancestries (paper). In a recent work we are showing that one can predict continental ancestry even from resting-state functional MRI. We have not established yet the reason for why this works, but we hypothesize that artifacts from head and brain morphology influence the observed connectivity pattern (paper). This highlights that neuroimaging data is susceptible to differences between genetic ancestries, i.e., population structure.
Do we need adjustment for population structure in neuroimaging? Yes, I think we should be more mindful about these effects and consider including genetic ancestry as covariates in imaging studies. The very same way we are addressing other confounding variables.