Comparative analysis of brain age prediction using structural and diffusion MRIs in neonates.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Brain age prediction using machine learning is a key biomarker for assessing neurodevelopment.
• Structural magnetic resonance imaging (sMRI) and diffusion magnetic resonance imaging (dMRI) are primary modalities, but their distinct contributions to infant age prediction are understudied.

Purpose of the Study:

• To investigate the differential impact of sMRI (T2w) and dMRI (FA) on predicting infant brain age.
• To explore how factors like prematurity, gender, and hemispherical asymmetry influence age prediction across these modalities.

Main Methods:

• Machine learning models were used to predict infant age from T2-weighted (T2w) and fractional anisotropy (FA) MRI data.
• Mean absolute errors (MAE) were compared between T2w and FA predictions.
• Exploratory analyses identified key brain regions contributing to age prediction for each modality.
• Statistical tests examined the influence of prematurity, gender, and hemispherical asymmetry on prediction biases.

Main Results:

• Both T2w and FA images yielded similar mean absolute errors in infant age prediction.
• T2w images emphasized the cerebral cortex and ventricles, while FA images highlighted the cerebral cortex and white matter tracts.
• FA images showed significant age prediction biases related to gender and hemispherical asymmetry (p<0.05), unlike T2w images.

Conclusions:

• T2w and FA MRI offer complementary information for infant brain age prediction, reflecting macrostructural and microstructural development, respectively.
• FA-based age prediction is more sensitive to variations in gender and brain asymmetry, suggesting distinct developmental trajectories.
• These findings provide novel insights into infant neurodevelopmental assessment and tracking.