Related Experiment Video
Updated: Jan 10, 2026
10:14
3D Scanning Technology Bridging Microcircuits and Macroscale Brain Images in 3D Novel Embedding Overlapping Protocol
Published on: May 12, 2019
7.6K
NeuroAgeFusionNet an ensemble deep learning framework integrating CNN, transformers, and GNN for robust brain age
Malla Sowmya1, Sushama Rani Dutta2
1Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Hyderabad, 500075, Telangana, India. somu42@gmail.com.
Scientific Reports
|November 25, 2025
Summary
NeuroAgeFusionNet accurately predicts brain age using a hybrid deep learning model. This novel approach enhances neuroimaging analysis for early disease detection and cognitive health assessment.
Area of Science:
- Neuroimaging and computational neuroscience.
- Artificial intelligence in medical diagnostics.
Background:
- Brain age prediction from MRI is vital for neurodegenerative disease diagnosis and cognitive health assessment.
- Conventional machine learning and existing deep learning models struggle with capturing complex MRI data, limiting accuracy and reliability.
- Current methods often fail to integrate spatial, contextual, and structural brain information effectively.
Purpose of the Study:
- To introduce NeuroAgeFusionNet, a novel hybrid deep learning framework for improved brain age estimation.
- To enhance the comprehensive feature representation of MRI data by optimizing spatial, contextual, and structural information.
- To increase the robustness and reliability of brain age predictions through an integrated uncertainty quantification module.
Main Methods:
- Developed a hybrid deep learning framework combining Convolutional Neural Networks (CNNs), Transformers, and Graph Neural Networks (GNNs).
- Implemented a feature fusion mechanism to optimize spatial, contextual, and structural features for a holistic data representation.
- Integrated an uncertainty quantification module to ensure prediction reliability and mitigate unreliable estimates.
Main Results:
- Achieved state-of-the-art performance on the UK Biobank dataset with a Mean Absolute Error (MAE) of 2.30.
- Demonstrated high accuracy with a Pearson correlation of 0.97 and an R-squared (R²) score of 0.96.
- Significantly outperformed conventional machine learning and existing deep learning approaches in brain age prediction.
Conclusions:
- NeuroAgeFusionNet offers a significant advancement in brain age estimation, providing high accuracy and reliability.
- The framework's ability to integrate diverse data features and quantify uncertainty makes it a valuable tool for clinical neuroscience.
- This approach holds promise for improved brain aging monitoring and the early detection of neurodegenerative diseases.
