3DCNN predicting brain age using diffusion tensor imaging.

Area of Science:

• Neuroimaging
• Deep Learning
• Computational Neuroscience

Background:

• Brain age prediction is a growing field in neuroimaging, offering insights into brain health.
• Diffusion tensor imaging (DTI) data has been underutilized for brain age estimation.
• Deep learning models show promise for complex neuroimaging analyses.

Purpose of the Study:

• To develop and validate a deep learning model for brain age prediction using DTI data.
• To identify specific brain regions and fiber tracts crucial for accurate age estimation.
• To investigate age-related alterations in brain axonal structure.

Main Methods:

• A 3D convolutional neural network (3DCNN) was developed for brain age prediction.
• The model was trained on fractional anisotropy (FA) data from 2406 participants (ages 17-60) across six datasets.
• A nested cross-validation strategy (10-fold CV) was implemented for robust performance evaluation.

Main Results:

• The 3DCNN model achieved a mean absolute error (MAE) of 2.785 and a correlation coefficient of 0.932.
• Grad-Cam++ visualization highlighted salient fiber tracts, including the genu of the corpus callosum and left cerebellar peduncle.
• The model demonstrated reliable brain age prediction performance.

Conclusions:

• The proposed 3DCNN model effectively predicts brain age using DTI-derived FA data.
• The study provides novel insights into age-related changes in the brain's white matter microstructure.
• Key fiber tracts identified by the model are critical for understanding brain aging processes.