Robust Brain Age Estimation Based on sMRI via Nonlinear Age-Adaptive Ensemble Learning.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Accurate brain age prediction is crucial for mental health prognosis and clinical trials.
• Discrepancies between chronological and brain age can indicate neurological health issues.
• Single machine learning models struggle with the diverse brain aging patterns.

Purpose of the Study:

• To develop a novel, nonlinearly adaptive ensemble model for enhanced brain age prediction.
• To integrate multiple machine learning algorithms for improved accuracy over single models.
• To provide a more reliable tool for assessing brain health and identifying potential risks.

Main Methods:

• An ensemble model combining Support Vector Regression (SVR), Convolutional Neural Network (CNN), GoogLeNet, and ResNet was developed.
• The model adaptively integrates individual algorithm predictions using age-dependent nonlinear weights.
• Model performance was evaluated using Mean Absolute Error (MAE) and Spearman correlation on the PAC 2019 dataset.

Main Results:

• The age-adaptive ensemble model achieved a Mean Absolute Error (MAE) of 3.19 on the PAC 2019 dataset.
• This result represents a significant improvement in accuracy for brain age estimation.
• The model demonstrated superior performance compared to single-algorithm approaches.

Conclusions:

• The proposed age-adaptive ensemble model offers a substantial advancement in brain age prediction accuracy.
• This improved accuracy can aid clinicians in identifying brain disease risks more effectively.
• The model holds potential as a valuable tool for brain science research and pharmaceutical development.