Related Experiment Video
Updated: Mar 14, 2026
05:55
Modeling the Functional Network for Spatial Navigation in the Human Brain
Published on: October 13, 2023
1.6K
BrainNetCNN: Convolutional neural networks for brain networks; towards predicting neurodevelopment
Jeremy Kawahara1, Colin J Brown1, Steven P Miller2
1Medical Image Analysis Lab, Simon Fraser University, Burnaby, BC, Canada.
Neuroimage
|October 4, 2016
Summary
BrainNetCNN, a novel convolutional neural network, accurately predicts neurodevelopmental outcomes in preterm infants using brain networks. This framework leverages topological features for enhanced predictive power in developmental assessments.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Medical Imaging
Background:
- Predicting neurodevelopmental outcomes in preterm infants is crucial for early intervention.
- Traditional Convolutional Neural Networks (CNNs) are limited in analyzing complex brain network topology.
Purpose of the Study:
- To introduce BrainNetCNN, a novel CNN framework designed for predicting clinical neurodevelopmental outcomes from structural brain networks.
- To evaluate BrainNetCNN's performance against existing methods using data from preterm infants.
Main Methods:
- Developed BrainNetCNN with novel edge-to-edge, edge-to-node, and node-to-graph convolutional filters to capture topological locality.
- Constructed structural brain connectivity networks from Diffusion Tensor Images (DTI) of preterm infants (27-46 weeks gestational age).
- Validated BrainNetCNN on synthetic networks and applied it to predict Bayley-III cognitive and motor scores at 18 months.
Main Results:
- BrainNetCNN outperformed a fully connected neural network on synthetic networks with simulated injuries.
- The framework demonstrated superior performance in predicting joint cognitive and motor scores compared to other methods.
- BrainNetCNN accurately estimated postmenstrual age within approximately two weeks and identified important brain connections for prediction.
Conclusions:
- BrainNetCNN offers a powerful new approach for analyzing brain network topology to predict neurodevelopmental outcomes in preterm infants.
- The method provides insights into the developing preterm infant brain by highlighting key connectivity patterns related to developmental trajectories.
