Machine-learning to characterise neonatal functional connectivity in the preterm brain.

Area of Science:

• Neuroscience
• Developmental Biology
• Medical Imaging

Background:

• Preterm birth significantly impacts brain development, leading to structural changes detectable by term-equivalent age.
• While structural alterations are known, the functional consequences of preterm birth on brain connectivity remain less understood.

Purpose of the Study:

• To investigate alterations in whole-brain functional connectivity in preterm infants compared to term-born neonates.
• To test the hypothesis that preterm birth causes specific functional connectivity changes by term-equivalent age.

Main Methods:

• Utilized machine learning, including group-independent component analysis and graphical lasso, to analyze functional connectivity networks.
• Employed random forest and nonlinear support vector machine for feature selection and subject classification.
• Analyzed data from 105 preterm infants and 26 term-born controls.

Main Results:

• Achieved 80% cross-validated classification accuracy in distinguishing preterm from term infants based on functional connectivity.
• Identified specific discriminative functional connectivity edges, primarily stronger in term neonates.
• Found that half of these discriminative edges involved the basal ganglia, highlighting subcortical alterations.

Conclusions:

• Preterm birth leads to significant alterations in brain functional connectivity by term-equivalent age.
• These findings confirm and specify connectivity changes between subcortical structures and cortical regions in preterm infants.