Predicting brain age with complex networks: From adolescence to adulthood

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Machine Learning

Background:

• Machine learning and deep learning show promise for predicting brain age from MRI scans.
• Accurate brain age prediction is crucial for understanding neurodevelopment and aging.

Purpose of the Study:

• To propose a novel complex network approach for predicting brain age using T1-weighted MRI scans.
• To characterize brain connectivity using centrality measures and deep neural networks for age prediction.

Main Methods:

• A structural connectivity model was created by dividing MRI scans into boxes and calculating Pearson's correlation.
• Brain connectivity was analyzed using centrality measures.
• A deep neural network was employed to predict brain age based on connectivity features.

Main Results:

• The approach achieved high accuracy with a correlation of r=0.89 and Mean Absolute Error (MAE) of 2.19 years on the primary dataset.
• On an independent test set, the MAE was 2.52 years, demonstrating robustness across different scanners and protocols.
• The method requires minimal imaging analysis (brain extraction, linear registration), ensuring computational efficiency.

Conclusions:

• The proposed complex network model offers an accurate, robust, and computationally efficient method for brain age prediction.
• This approach provides novel insights into brain aging patterns and identifies specific anatomical regions affected by aging.
• The framework's simplicity and efficiency make it suitable for large and heterogeneous neuroimaging datasets.