Five Best Practices for fMRI research

The replication crisis in science has not spared functional magnetic resonance imaging (fMRI) research. A range of issues including insufficient control false positives, code bugs, concern regarding generalizability and replicability findings, inadequate characterization physiological confounds, over-mining repository datasets, the small sample sizes/low power many early studies have led to hearty debate both field press about usefulness viability fMRI. Others still see enormous potential for fMRI diagnosing conditions that do otherwise lend themselves non-invasive biological measurement, from chronic pain neurological psychiatric illness. How we reconcile limitations with hype over its potential? Despite papers hailed by as nail coffin fMRI, dead salmon incident 2009 cluster failure more recently, funders, researchers, general public seem reduced their appetite pictures brain maps, or gadgets word “neuro” name. Multiple blogs exist sole purpose criticizing such enterprise. should certainly give ‘neuroimagers’ pause, reason soul-search. It is important than ever clarify when it useful. method remains best noninvasive tool research questions, however imperfect imprecise may be. However, address past limitations, I argue neuroimaging researchers planning future need consider following five factors: power/effect size, design optimization, replicability, data sharing.