Related Experiment Video
Updated: Dec 17, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
16.0K
[Single-modal neuroimaging computer aided diagnosis for schizophrenia based on ensemble learning using privileged
Lu Shen1, Qianting Wang1, Jun Shi1
1Shanghai Institute for Advanced Communication and Data Science, School of Communication and Information Engineering, Shanghai University, Shanghai 200444, P.R.China.
Summary
A new ensemble learning algorithm enhances schizophrenia diagnosis using single neuroimaging data. This method improves classification accuracy, sensitivity, and specificity for both structural and functional MRI scans.
Area of Science:
- Neuroimaging and Artificial Intelligence
- Computational Psychiatry
Background:
- Neuroimaging plays a crucial role in the computer-aided diagnosis (CAD) of schizophrenia.
- Existing single-modal neuroimaging CAD systems for schizophrenia often face performance limitations.
- Traditional Learning Using Privileged Information (LUPI) methods typically require additional data modalities.
Purpose of the Study:
- To propose an improved ensemble learning algorithm for schizophrenia diagnosis using single neuroimaging data.
- To overcome the limitations of traditional LUPI methods by not requiring extra privileged information modalities.
- To enhance the classification performance of schizophrenia detection from neuroimaging datasets.
Main Methods:
- An ensemble learning algorithm based on Learning Using Privileged Information (LUPI) was developed.
- Extreme Learning Machine based Auto-Encoder (ELM-AE) was used for feature representation learning from single-modal neuroimaging data.
- Random projection generated feature subspaces, followed by training multiple Support Vector Machine plus (SVM+) classifiers, and combining them into a strong classifier.
Main Results:
- The proposed algorithm demonstrated superior diagnostic performance on a public schizophrenia neuroimaging dataset.
- On structural MRI (sMRI) data, classification accuracy, sensitivity, and specificity were 72.12% ± 8.20%, 73.50% ± 15.44%, and 70.93% ± 12.93%, respectively.
- On functional MRI (fMRI) data, classification accuracy, sensitivity, and specificity were 72.33% ± 8.95%, 68.50% ± 16.58%, and 75.73% ± 16.10%, respectively.
Conclusions:
- The proposed LUPI-based ensemble learning algorithm effectively improves schizophrenia diagnosis using single-modal neuroimaging data.
- The method successfully addresses the need for additional privileged information, allowing direct application to single-modal data.
- This approach offers a promising tool for enhancing schizophrenia detection and suggests broader applications in medical image analysis.
