Brain Age Prediction of Children Using Routine Brain MR Images via Deep Learning

Area of Science:

• Pediatric neuroimaging
• Artificial intelligence in medicine
• Developmental neuroscience

Background:

• Accurate brain age prediction is crucial for pediatric development analysis and disease diagnosis.
• Traditional methods using 3D MRI (T1WI, DTI) require complex preprocessing and extra scanning time, limiting clinical utility in children.
• Routine brain MRI is more accessible and less burdensome for pediatric patients.

Purpose of the Study:

• To develop and validate an end-to-end deep learning system for predicting brain age in children using routine 2D brain MR T1-weighted images.
• To establish a quantitative and accurate method for assessing brain development in children aged 0-5 years.
• To overcome the limitations of traditional brain age estimation techniques in pediatric populations.

Main Methods:

• A dataset of 220 stacked 2D routine clinical brain MR T1-weighted images from healthy children (0-5 years) was utilized, with 176 for training and 44 for testing.
• Data augmentation techniques including scaling, rotation, translation, and gamma correction were applied to enhance the training dataset.
• A 10-layer 3D convolutional neural network (CNN) was designed and implemented for brain age prediction.

Main Results:

• The 3D CNN model achieved high accuracy in predicting brain age, with a mean absolute deviation (MAE) of 67.6 days and a correlation coefficient (R) of 0.985 on test data.
• Performance was particularly strong for children under 2 years old (MAE: 28.9 days, R: 0.983), outperforming predictions for older children (MAE: 110.0 days, R: 0.883).
• The model demonstrated a coefficient of determination (R²) of 0.971 overall, indicating a robust fit to the data.

Conclusions:

• The proposed deep learning system offers a reliable and efficient method for predicting brain age in young children using routine MR imaging.
• This AI-driven approach can significantly aid in the early detection of developmental abnormalities and neurological disorders.
• The system's superior performance in younger children highlights its potential for sensitive monitoring of early brain development.