The Identification of Early Neuroimaging Biomarkers of (Impaired) Language Development in Extremely and Very Preterm Children

Abstract

Background: Prematurity is associated with a diverse range of neurodevelopmental sequelae. For instance, a premature birth can lead to language impairment. Various studies have investigated the language development of extremely and very preterm children. However, their language development has barely been correlated with brain characteristics, while brain abnormalities could have prognostic value for the identification of children at risk for a language delay. This scoping review aims to synthesize the current literature by providing an overview of studies which investigate neuroimaging biomarkers of impaired language development in extremely and very preterm children. This could stimulate the timely start of interventions that attenuate the negative impact of prematurity on language skills, when neuroplasticity is at its peak.

Methods: We selected a subset of articles by conducting an advanced PubMed search and screening their abstracts and titles. Studies were eligible for inclusion when the study consisted of a population of extremely or very preterm infants (equal to or less than 32 weeks GA), the study used a neuroimaging method at term-equivalent age to measure brain abnormalities or specific parameters, the study reported at least one language outcome, and a correlational measure could be extracted that represented an association between brain and language. In total, 28 studies were included.

Results: The results were split between extremely and very preterm infants: seven studies were conducted on extremely preterm infants, while 21 studies examined very preterm infants. Overall, no major differences were found between the two groups. Most significant correlations between brain metrics and language were found for biomarkers from the categories brain morphometry and brain volumes. White matter had the most predictive power from all tissue types. Studies that adopted a novel approach by using an innovative neuroimaging method and/or multimodal research design were among the most successful ones in classifying preterm children at high or low risk for a language delay based on their brain features. The results were quite inconsistent for biomarkers from the categories structural abnormalities and white matter microstructure and connectivity.

Conclusion: Despite some inconsistency within the results, various neuroimaging biomarkers were discovered that have added value for the identification of infants with an increased risk for language impairment. Brain morphometric and volumetric biomarkers were the most promising, whereas biomarkers from the categories structural abnormalities and white matter microstructure and connectivity were not reliable enough. Further research is needed to confirm this. Future studies are advised to use stricter inclusion criteria for more selective data collection, to facilitate comparisons between included articles. Future studies should also focus on the potential of deep learning (multimodal) models and go beyond the golden standard MRI.