Characterising longitudinal dynamics of infant neurodevelopment using functional change point detection

Habituation is an early-developing cognitive process linked to early learning whereby irrelevant stimuli are filtered out; it can be assessed using neuroimaging tools, including functional near-infrared spectroscopy (fNIRS). However, conventional fNIRS analysis approaches typically assume linear time invariance and analyse summary statistics of the data, potentially limiting insight into the nature of the changing evoked haemodynamic response. Functional data analysis (FDA) is a branch of statistics which treats entire mathematical functions as the unit of analysis rather than discrete points; within this framework, functional change point (FCPt) detection can identify significant changes in data structure, such as the mean haemodynamic response function. Using infant fNIRS data from a rural Gambian population (n= 204) recorded using a habituation and novelty detection paradigm, the use of functional change point detection with wild binary segmentation was implemented to identify structural changes in the evoked haemodynamic response. Group-level analysis from participants assessed at 5-, 8- and 12mo of age identified structural change corresponding to the evoked haemodynamic response in locations consistent with previous analyses. Further analyses indicated consistent habituation at 8- and 12mo and an age-related shift in the timing of decreasing haemodynamic responses, suggesting older infants in the first year of life habituate more rapidly to auditory stimuli. These results provide novel evidence to support the use of FDA and FCPt detection in the fNIRS field more widely, as well as improving understanding of the development of habituation during early infancy.