Human-to-monkey transfer learning identifies the frontal white matter as a key determinant for predicting monkey...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Imaging

Background:

• Brain age estimation using AI offers an objective view of aging.
• Deep learning (DL) for brain age prediction faces challenges in sample size and interpretability.
• Animal models, like rhesus monkeys, are valuable for studying aging due to controlled factors.

Purpose of the Study:

• To adapt AI-driven brain age estimation for animal models, specifically rhesus monkeys.
• To overcome data insufficiency in animal brain MRI datasets using transfer learning.
• To develop an interpretable and accurate AI model for non-human primate brain aging research.

Main Methods:

• Utilized transfer learning by applying AI models pre-trained on 8,859 human brain MRIs to monkey brain MRIs.
• Evaluated 3D ResNet and 2D global-local transformer models for brain age estimation.
• Identified key brain regions contributing to age prediction through model analysis.

Main Results:

• Transfer learning significantly improved the accuracy and stability of monkey brain age estimation.
• The 3D ResNet model achieved the highest accuracy (MAE = 1.83 years), followed by the 2D global-local transformer (MAE = 1.92 years).
• Frontal white matter was identified as the most critical feature for predicting monkey brain age.

Conclusions:

• Transfer learning effectively addresses data limitations in applying DL to animal brain aging.
• This study presents the first interpretable, AI-based brain age estimator for non-human primates.
• The developed method has broad potential for AI applications in diverse animal and disease models.