Related Experiment Video
Updated: Aug 1, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Progressive Graph-Based Transductive Learning for Multi-modal Classification of Brain Disorder Disease
Zhengxia Wang1, Xiaofeng Zhu2, Ehsan Adeli2
1Department of Radiology and BRIC, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA; Department of Information Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China.
Medical Image Computing and Computer-Assisted Intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
|April 8, 2017
Summary
This study introduces progressive Graph-based Transductive Learning (pGTL) for improved computer-assisted diagnosis. The novel method enhances classification accuracy by iteratively refining data representations and relationships, outperforming conventional approaches.
Area of Science:
- Medical imaging analysis
- Machine learning for healthcare
- Computational neuroscience
Background:
- Graph-based Transductive Learning (GTL) aids diagnosis with limited data.
- Conventional GTL graphs built in the feature domain may not optimize label domain classification.
- This limitation can reduce diagnostic accuracy.
Purpose of the Study:
- To develop a progressive Graph-based Transductive Learning (pGTL) method for enhanced diagnostic classification.
- To improve the intrinsic data representation by iteratively refining feature and label domains.
- To extend pGTL for multi-modal imaging data integration.
Main Methods:
- pGTL iteratively refines subject relationships using label domain representations.
- Intrinsic data representation is updated based on refined relationships.
- Verification on training data ensures optimal classification for new data.
- Multi-modal imaging data (MRI, PET) integration was explored.
Main Results:
- The pGTL method demonstrated promising classification results.
- Improved accuracy and robustness were achieved, particularly with multi-modal data.
- Effective identification of Alzheimer's disease (AD), Mild Cognitive Impairment (MCI), and Normal Control (NC) subjects.
Conclusions:
- The proposed pGTL method offers a significant advancement over conventional GTL.
- Iterative refinement of data representation and multi-modal data integration enhance diagnostic performance.
- pGTL shows strong potential for computer-assisted diagnosis in neurodegenerative diseases.
