BRAIN AGE PREDICTION BASED ON RESTING-STATE FUNCTIONAL CONNECTIVITY PATTERNS USING CONVOLUTIONAL NEURAL NETWORKS.

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Brain age prediction using neuroimaging aids in understanding typical development and neuropsychiatric disorders.
• Resting-state fMRI (rsfMRI) and functional connectivity (FC) are established tools for brain age prediction.
• Existing methods often use coarse-grained FC, potentially losing detailed information.

Purpose of the Study:

• To develop a deep learning model utilizing fine-grained, whole-brain voxel-wise FC measures from rsfMRI for improved brain age prediction.
• To investigate the efficacy of convolutional neural networks (CNNs) in extracting informative features from detailed FC data.

Main Methods:

• A deep learning approach employing CNNs was developed.
• The model was trained on whole-brain voxel-wise FC measures derived from rsfMRI data.
• Performance was evaluated on a large-scale rsfMRI dataset.

Main Results:

• The deep learning model demonstrated superior performance in predicting brain age when using fine-grained FC measures.
• CNNs effectively learned relevant features from the detailed connectivity data.

Conclusions:

• Fine-grained whole-brain FC measures, when processed by deep learning models like CNNs, significantly enhance brain age prediction accuracy.
• This approach offers a more nuanced understanding of brain maturation and potential deviations in neuropsychiatric conditions.