MRI-based brain age prediction model for children under 3 years old using deep residual network

Area of Science:

• Neuroscience
• Pediatric Imaging
• Artificial Intelligence in Medicine

Background:

• Early identification of abnormal brain development in children under 3 is crucial.
• Existing brain age prediction models are primarily designed for adults, limiting their use in pediatric populations.
• Brain biological age, predicted from neuroimaging, correlates with health and cognitive abilities.

Purpose of the Study:

• To develop and validate an accurate brain age prediction model for children aged 0-3 years using deep learning.
• To assess the model's performance in capturing age-related brain changes in early childhood.
• To investigate the utility of the model in identifying developmental differences in specific pediatric conditions.

Main Methods:

• Collected 658 T1-weighted MRI scans from healthy children aged 0-3 years.
• Developed a deep learning model, specifically a convolutional neural network (CNN), for brain age prediction.
• Validated the model's accuracy by comparing predicted brain age with chronological age, achieving a 91% correlation.

Main Results:

• The deep learning model achieved high accuracy in predicting chronological age from T1-weighted MRI scans in young children.
• The model's performance was comparable to Support Vector Regression (SVR) methods.
• Abnormal brain age predictions were observed in children with extremely low birth weight (delayed) and Attention-Deficit/Hyperactivity Disorder (ADHD) (accelerated).

Conclusions:

• A child-specific deep learning model accurately predicts brain age using minimally processed MRI data.
• This model serves as a valuable quantitative tool for detecting abnormal brain development in early childhood.
• The findings support the early identification and intervention of neurodevelopmental disorders through advanced neuroimaging analysis.