Systematic exploration of task-based functional connectivity analyses for infant fNIRS data: Toward single trial measures of connectivity

Significance: There is increasing interest in task-based functional connectivity analyses to examine the emergence of functional networks during specific cognitive states starting early in development (e.g., infancy). Studies must be carefully designed to collect sufficient high-quality neuroimaging data, while remaining sensitive to infants’ developing motor and cognitive abilities. This presents challenges in task design and limits suitable task-based functional connectivity analysis techniques for infant data.Aim: This study aims to expand the established analysis approaches for infant functional connectivity studies through re-analysis of an existing dataset with a recognised signature of task-based functional connectivity. We assess the feasibility of moving away from analyses that require several contiguous trials for each condition, a highly unusual feature of task-based fNIRS datasets.Approach: Background functional connectivity analyses utilise the residual neural response (i.e., what remains after removing task-evoked responses) to examine task-based functional connectivity that occurs in the “background”. We systematically examine the viability of three analytic approaches – downsampling, shuffling and individual trial – to push the boundaries of background functional connectivity analyses for infant fNIRS data.Results: Downsampling was informative about the number of trials needed for well-powered analyses and the timing of the functional connectivity effect. Removing trials from the start of the block was not viable, but this approach was viable when trials were removed from the end. The shuffling approach, which randomly reorders trial-level residuals within conditions, was not viable. Results indicated that BGFC analyses could be applied at the single trial level, however, averaging trial-level residuals before performing the correlational analyses was the more viable approach.Conclusions: This systematic exploration indicates that downsampling (from the end) and single trial connectivity are viable BGFC analysis approaches. Results will benefit the field by informing task-design decisions and increasing the accessibility of BGFC analyses for infant task-based functional connectivity studies.