A novel deep learning-based brain age prediction framework for routine clinical MRI scans.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Gerontology

Background:

• Physiological brain aging correlates with cognitive decline and structural brain changes.
• Brain age prediction using routine 2D MRI scans is challenging and underexplored.
• Existing methods often struggle with the accuracy and speed required for clinical application.

Purpose of the Study:

• To develop and validate a novel deep learning framework for accurate brain age prediction from clinical 2D T1-weighted MRI scans.
• To assess the framework's performance on cognitively unimpaired individuals and its association with neurodegenerative diseases.
• To explore the potential of this framework for routine clinical examinations and early detection of brain abnormalities.

Main Methods:

• A deep learning model was trained using a large dataset of 3D MRI scans (N=8681).
• The model was applied to predict brain age from 2D MRI scans of cognitively unimpaired (CU) subjects (N=175).
• Age bias correction was applied, and performance was evaluated using Mean Absolute Error (MAE) and Pearson correlation (r).

Main Results:

• The framework achieved accurate and rapid brain age prediction on clinical 2D MRI scans from CU subjects (MAE=2.73 years, r=0.918).
• A significantly greater brain age gap was observed in subjects with Alzheimer's disease (AD) compared to CU subjects (p<0.001).
• Increased brain age gap correlated with disease progression in both AD (p<0.05) and Parkinson's disease (p<0.01).

Conclusions:

• The developed deep learning framework offers a robust method for brain age prediction using readily available 2D clinical MRI scans.
• The brain age gap serves as a potential imaging biomarker for neurodegenerative diseases like AD and Parkinson's.
• This framework holds promise for integration into routine clinical practice to aid in early detection and improve patient outcomes.