Related Experiment Video
Updated: May 16, 2025
Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example
Published on: October 24, 2012
Enhancing brain age estimation under uncertainty: A spectral-normalized neural gaussian process approach utilizing
Zeqiang Linli1, Xingcheng Liang2, Zhenhua Zhang2
1School of Mathematics and Statistics, Guangdong University of Foreign Studies, Guangzhou, 510420, PR China; Laboratory of Language Engineering and Computing, Guangdong University of Foreign Studies, 510420, Guangzhou, PR China; MOE-LCSM, School of Mathematics and Statistics, Hunan Normal University, Changsha, 410006, PR China.
This study introduces a novel deep learning method, Spectral-normalized Neural Gaussian Process (SNGP), for estimating brain age with uncertainty. The 2.5D approach enhances clinical applications by improving accuracy and integrating dimensional data cost-effectively.
Area of Science:
- Neuroimaging
- Artificial Intelligence
- Biomarker Development
Background:
- Brain age gap is a key biomarker for detecting brain abnormalities.
- Deep learning models for brain age estimation lack uncertainty quantification, posing clinical risks.
- Existing 3D models are complex, and 2D approaches limit data integration.
Purpose of the Study:
- To develop a cost-effective deep learning method for brain age estimation with uncertainty.
- To integrate dimensional data and uncertainty estimation seamlessly within a single network.
- To compare deep learning methods for brain age uncertainty estimation using Pearson correlation coefficient.
Main Methods:
- Introduced Spectral-normalized Neural Gaussian Process (SNGP) with a 2.5D slice approach.
- Utilized 11 public datasets (N=6327) for training and an independent dataset (N=301) for validation.
- Conducted five controlled experiments to validate the SNGP method's performance and capabilities.
Main Results:
- SNGP demonstrated excellent uncertainty estimation and generalization (MAE=2.95 on public, MAE=3.47 on independent datasets).
- Uncertainty adjustment improved accelerated brain aging detection in ADHD adolescents by 38%.
- The 2.5D approach outperformed 2D methods, and SNGP showed Out-of-Distribution detection capabilities.
Conclusions:
- SNGP offers a cost-effective method for brain age estimation with uncertainty using 2.5D slicing.
- The method enhances clinical applications by improving performance and data integration without added complexity.
- The study validates SNGP's robustness, generalization, and potential for detecting neurodevelopmental and aging-related abnormalities.
