Predicting 2-year neurodevelopmental outcomes in preterm infants using multimodal structural brain magnetic resonance...

Area of Science:

• Neuroscience
• Developmental Pediatrics
• Medical Imaging

Background:

• Neurodevelopmental outcomes in preterm infants vary with gestational age.
• Early prediction of these outcomes is crucial for timely intervention.
• Brain structural networks offer potential biomarkers for neurodevelopmental trajectories.

Purpose of the Study:

• To explore brain structural networks in extremely preterm (EP) and very-to-late preterm (V-LP) infants.
• To predict 2-year neurodevelopmental outcomes using machine learning models.
• To identify key neuroimaging features, particularly local connectivity, for outcome prediction.

Main Methods:

• Utilized MRI and diffusion MRI on 62 EP and 131 V-LP infants at term-equivalent age.
• Developed multimodal feature sets for volumetric and structural network analysis.
• Employed linear and nonlinear machine learning models to predict Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III) scores.

Main Results:

• Models incorporating local connectivity features showed high predictive performance for BSID-III scores.
• Local connectivity features predicted cognitive scores in preterm (17% variance) and V-LP (17%) infants.
• Local connectivity predicted motor scores in EP infants (15% variance) and language scores in preterm infants (15% variance).

Conclusions:

• Multimodal feature sets, especially local connectivity, are valuable for predicting neurodevelopmental outcomes in preterm infants.
• Machine learning effectively utilizes neuroimaging data to understand microstructural changes.
• These findings support early intervention strategies informed by neuroimaging biomarkers.