Can we predict subject-specific dynamic cortical thickness maps during infancy from birth?

Area of Science:

• Neuroscience
• Developmental Neuroscience
• Medical Imaging

Background:

• Early human cerebral cortex development is complex and not well understood.
• Cortical thickness is a key morphological feature, sensitive to neurodevelopment and disorders.
• Early postnatal cortical development is largely unexplored.

Purpose of the Study:

• To determine if neonatal MRI data can predict future dynamic cortical thickness development in infants.
• To establish a model for understanding early brain development.
• To enable early detection of abnormal brain development.

Main Methods:

• Developed a novel learning-based method: Dynamically-Assembled Regression Forest (DARF).
• Used neonatal MRI features to predict cortical thickness maps.
• Applied the method to 15 healthy infants, predicting maps at 3, 6, 9, and 12 months.

Main Results:

• Achieved mean absolute errors of 0.209 mm (3 mo), 0.332 mm (6 mo), 0.340 mm (9 mo), and 0.321 mm (12 mo).
• Demonstrated region-specific prediction precision, higher in unimodal cortex than association cortex.
• Showed that incorporating more early time points improves prediction accuracy.

Conclusions:

• The DARF method can effectively predict infant cortical thickness development from neonatal MRI.
• Prediction accuracy varies by cortical region, reflecting differential developmental trajectories.
• Early MRI data holds significant potential for modeling and monitoring infant brain development.