BRAIN AGE ESTIMATION USING LSTM ON CHILDREN'S BRAIN MRI.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Pediatric Neurology

Background:

• Children's brain MRI analysis is crucial for understanding brain health and development.
• Brain age prediction serves as a key biomarker in this field.
• Accurate brain age estimation aids in identifying developmental abnormalities.

Purpose of the Study:

• To propose a novel recurrent neural network framework for enhanced brain age estimation in children using 3D brain MRI data.
• To evaluate the efficacy of the proposed 2D-ResNet18+LSTM method against existing 3D neural network approaches.
• To establish a more accurate biomarker for pediatric brain development and health.

Main Methods:

• The proposed method, 2D-ResNet18+LSTM, treats 3D brain MRI volumes as sequences of 2D images.
• It employs 2D ResNet18 for feature extraction, followed by a pooling layer for feature reduction.
• A Long short-term memory (LSTM) layer and a final regression layer complete the network architecture.

Main Results:

• The 2D-ResNet18+LSTM method demonstrated superior performance in brain age estimation compared to traditional 3D neural networks.
• Validation on a public multisite NIH-PD dataset and a second multisite dataset confirmed the method's effectiveness.
• The framework achieved better results in predicting brain age, indicating improved accuracy.

Conclusions:

• The 2D-ResNet18+LSTM framework offers a more effective approach for brain age prediction in pediatric populations.
• This method provides a valuable tool for analyzing brain health and development using MRI data.
• The findings suggest a promising direction for advancing neurodevelopmental research through advanced AI techniques.