Related Experiment Video
Updated: May 4, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
15.2K
Identifying informative imaging biomarkers via tree structured sparse learning for AD diagnosis.
Manhua Liu1, Daoqiang Zhang, Dinggang Shen
1Department of Instrument Science and Engineering, SEIEE, Shanghai Jiao Tong University, Dong Chuan Rd #800, Shanghai, China, mhliu@sjtu.edu.cn.
Neuroinformatics
|December 17, 2013
Summary
This study introduces a novel tree-structured sparse learning method for Alzheimer's disease (AD) and mild cognitive impairment (MCI) neuroimaging analysis. The method effectively identifies key imaging biomarkers for improved disease classification and interpretation.
Area of Science:
- Neuroimaging
- Machine Learning
- Computational Neuroscience
Background:
- Neuroimaging is crucial for tracking Alzheimer's disease (AD) and mild cognitive impairment (MCI) progression and treatment effectiveness.
- Identifying relevant neuroimaging biomarkers is challenging due to complex, unknown patterns of structural degeneration.
- Existing sparse learning methods often overlook spatial structure, limiting biomarker identification.
Purpose of the Study:
- To propose a novel sparse learning method incorporating tree-structured regularization for Alzheimer's disease (AD) and mild cognitive impairment (MCI) neuroimaging.
- To effectively capture multi-scale pathological degeneration patterns and identify informative imaging biomarkers.
- To enhance disease classification and interpretation accuracy in neurodegenerative conditions.
Main Methods:
- Developed a new tree construction method using hierarchical agglomerative clustering of voxel-wise imaging features, considering spatial adjacency, similarity, and discriminability.
- Applied tree-structured regularization to sparse learning for capturing imaging structures and selecting relevant biomarkers.
- Utilized a support vector machine (SVM) classifier trained on selected features for disease classification.
Main Results:
- Achieved high classification accuracies: 90.2% for AD vs. Normal Controls (NC), 87.2% for progressive MCI (pMCI) vs. NC, and 70.7% for pMCI vs. stable MCI (sMCI).
- Demonstrated promising performance compared to existing state-of-the-art methods.
- Validated the method using baseline MR images from 830 subjects in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
Conclusions:
- The proposed tree-structured sparse learning method effectively identifies informative neuroimaging biomarkers for AD and MCI.
- This approach enhances the accuracy of disease classification and interpretation.
- The method shows significant potential for clinical applications in diagnosing and monitoring neurodegenerative diseases.
