GACT: A Two-Stage Age Prediction Model Combining a Global Attention Block.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Brain age estimation using neuroimaging is vital for understanding brain development and neurological diseases.
• Current deep learning models using functional MRI (fMRI) often rely on spatial maps or connectivity, potentially losing detailed brain information.

Purpose of the Study:

• To develop a novel deep learning approach for brain age prediction using unsegmented fMRI data.
• To better capture spatiotemporal information from fMRI data for improved age estimation.

Main Methods:

• Utilized raw, unsegmented fMRI data as input features.
• Integrated convolutional neural networks (CNNs) and transformer models to extract spatial and temporal features.
• Employed a Multi-Layer Perceptron (MLP) for the final age prediction.

Main Results:

• The proposed model demonstrated outstanding performance in brain age prediction tasks.
• An explainability method identified key brain regions influencing age regression.

Conclusions:

• Directly using unsegmented fMRI data with CNNs and transformers offers a powerful new methodology for brain age estimation.
• The findings provide valuable insights for future neuroimaging research and understanding brain aging.