Related Experiment Video
Updated: Sep 22, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
15.8K
Biomarkers identification for Schizophrenia via VAE and GSDAE-based data augmentation.
Qi Huang1, Chen Qiao1, Kaili Jing2
1School of Mathematics and Statistics, Xi'an Jiaotong University, Xi'an, 710049, China.
Computers in Biology and Medicine
|May 19, 2022
Summary
This study introduces GS-VDAE, a novel data augmentation method for high-dimensional, small-sample MRI data. GS-VDAE improves biomarker identification accuracy by generating more representative augmented samples, outperforming standard variational auto-encoders.
Area of Science:
- Medical Imaging Analysis
- Machine Learning
- Biomarker Discovery
Background:
- High-dimensional, small-sample size (HDSSS) MRI data presents challenges for biomarker identification.
- Deep learning methods struggle with HDSSS data limitations.
- Data augmentation is crucial for few-shot learning in such scenarios.
Purpose of the Study:
- To develop an advanced data augmentation technique for HDSSS MRI data.
- To enhance the accuracy of biomarker identification in medical datasets.
- To address the limitations of existing generative models in medical data analysis.
Main Methods:
- Proposed GS-VDAE, integrating Variational Auto-Encoder (VAE) with Graph Regularized Sparse Deep Autoencoder (GSDAE).
- Embedded data generation within GSDAE to leverage significant features from original samples.
- Utilized a regression feature selection method with truncated nuclear norm regularization for biomarker selection.
Main Results:
- GS-VDAE generated samples achieved a classification accuracy of 0.84, significantly higher than VAE's 0.74.
- Biomarker selection on schizophrenia data showed improved classification accuracy with fewer features using GS-VDAE augmented data.
- The model effectively ensures augmented samples retain significant characteristics of the original data.
Conclusions:
- GS-VDAE offers a superior approach to data augmentation for HDSSS MRI data.
- The method enhances biomarker identification accuracy and efficiency.
- This technique validates its effectiveness in real-world medical data, such as schizophrenia datasets.
