Neonatal Multisensory Processing in Preterm and Term Infants Predicts Sensory Reactivity and Internalizing Tendencies in Early Childhood

Abstract

Multisensory processes include the capacity to combine information from the different senses, often improving stimulus representations and behavior. The extent to which multisensory processes are an innate capacity or instead require experience with environmental stimuli remains debated. We addressed this knowledge gap by studying multisensory processes in prematurely born and full-term infants. We recorded 128-channel event-related potentials (ERPs) from a cohort of 55 full-term and 61 preterm neonates (at an equivalent gestational age) in response to auditory, somatosensory, and combined auditory-somatosensory multisensory stimuli. Data were analyzed within an electrical neuroimaging framework, involving unsupervised topographic clustering of the ERP data. Multisensory processing in full-term infants was characterized by a simple linear summation of responses to auditory and somatosensory stimuli alone, which furthermore shared common ERP topographic features. We refer to the ERP topography observed in full-term infants as "typical infantile processing" (TIP). In stark contrast, preterm infants exhibited non-linear responses and topographies less-often characterized by TIP; there were distinct patterns of ERP topographies to multisensory and summed unisensory conditions. We further observed that the better TIP characterized an infant's ERPs, independently of prematurity, the more typical was the score on the Infant/Toddler Sensory Profile (ITSP) at 12 months of age and the less likely was the child to the show internalizing tendencies at 24 months of age. Collectively, these results highlight striking differences in the brain's responses to multisensory stimuli in children born prematurely; differences that relate to later sensory and internalizing functions.

Keywords: EEG; Hypersensitivity; Infant; Multisensory; Neonate brain; Premature; Sound; Touch.

Fig. 1

Neonates were tested using a 128-electrode EEG net in their isolettes a if they were in the NICU (preterm) or in their bassinettes in the newborn nursery (full-term). Calibrated tactile stimulation was delivered via a custom 3D-printed holder fastened with a neoprene glove to the infant’s hand allowing the air pulse to always have the same pressure and target (b). The paradigm presents stimuli in random order and at random intertrial intervals, with a standard /ga/ sound and the tactile air puff stimulation presented separately for unisensory conditions or simultaneously for multisensory conditions (see c). Finally, in the analysis, the unisensory responses are algebraically added to produce a summed response and compared with the true multisensory response (panel c, lower)

Fig. 2

ERP analyses contrasting multisensory and summed unisensory responses from full term and preterm infants (panels a and b, respectively). The uppermost row displays group-averaged data from all electrodes for the multisensory pair and summed unisensory conditions. The next row displays group-averaged data (s.e.m. indicated by shaded area) from an exemplar fronto-central electrode (see inset). The next row displays the percentage of the electrode montage exhibiting a significant difference as a function of time. The bottom row displays the spatial correlation between multisensory and summed unisensory responses (blue line) as well as time periods of significant differences (orange areas)

Fig. 3

Template maps identified over the 176–231 ms post-stimulus period via unsupervised hierarchical clustering of the ERP topography using the group-averaged dataset concatenated across conditions and full-term and preterm infants. The maps are displayed with the nose upwards and left hemiscalp on the left. The bar graphs display the single-subject fitting based on the spatial correlation between the template maps shown the top row and each time point of the infant’s data over the 176–231 ms post-stimulus period. The bars show the average percentage of time each template map yielded the highest spatial correlation (error bars indicate s.e.m.)