Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Structural and Functional Brain Connectomics Predicts Language Outcomes in Moderate to Late Preterm and Early Term Infants
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
Xiujuan Geng1, Pegg HY Chan, Winnie CW Chu, Hugh S Lam, Patrick CM Wong; 1Chinese University of Hong Kong
Introduction Studies indicate that both structural and functional large-scale brain networks are present in infants born preterm and at term, while postnatal brain connectome undergoes profound re-organization. However, it remains unclear how structural and functional connectomic features are influenced by preterm birth and how they relate to later language outcomes in infants. Our recent work has shown that white matter integrity is significantly affected by preterm birth, and serves as a strong predictor of language outcomes. We further hypothesize that both structural and functional connectomic features are affected by preterm birth and are predictive to later language abilities in infants born preterm and early term. Methods Two hundred and thirty-four healthy Chinese MLPT and early term infants (mean gestation age at birth: 36.6 weeks, range: 32.0-40.2 weeks) were recruited. Subjects were scanned at approximate two months of chronological age (mean=9.7, range=3.0-18.6 weeks) during natural sleep on a Siemens Magnetom Prisma 3T scanner. Diffusion-weighted images (DWI) and resting-state fMRI were collected. Language ability, including expressive, receptive and composite scores was assessed at 12 and 24 months using Bayley Scales of Infant and Toddler Development III. Fiber tractography was performed on preprocessed DWI and structural connectivity matrices were generated using the Brainnetome atlas to extract FA values along tracts connecting each pair of ROIs. Functional connectivity matrices were estimated by Pearson correlations between preprocessed fMRI time courses from each pair of ROIs. For both connectivity types, nodal degree (for each node) and network-level connections (based on the network atlas) were extracted and used as input features to classify language outcomes (median split to high and low) measured at 1 and 2 years of age. A support vector machine with nested K-fold cross validation was employed to predict language outcomes. Results Controlling for sex, age, and parent education using GLM, we found significant preterm effects on most structural nodal degrees but not on functional ones. Preterm birth impacts structural networks more. Prediction results show that both structural and functional metrics significantly predict language outcomes at 1 and 2 years, with sensitivity and AUC ranging from 0.76 to 0.93. The most predictive structural nodal degrees include the bilateral middle, inferior, and superior temporal cortices, superior frontal gyrus, and occipital gyrus. The most predictive functional nodal degrees involve the bilateral superior and middle temporal cortices and postcentral gyri. For network-level metrics, key structural connections include limbic-visual, limbic-motor, limbic-dorsal attention, and within the limbic network, while top functional connections are found within motor, limbic, control, and auditory networks. Conclusion We found that structural connectomic features, rather than functional ones, are significantly affected by preterm birth. Both structural and functional connectome metrics predict language outcomes up to two years of age. Specifically, nodal degrees from the bilateral temporal cortices are the most predictive, while distinct structural and functional network profiles also contribute. Our findings suggest that rather than language-speific regions, the early development of distributed networks, spanning limbic, attention, and control systems, may play a crucial role in later language outcomes.
Topic Areas: Language Development/Acquisition,