Related Experiment Video
Updated: May 13, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Semi-supervised multimodal relevance vector regression improves cognitive performance estimation from imaging and
Bo Cheng1, Daoqiang Zhang, Songcan Chen
1Department of Computer Science and Engineering, Nanjing University of Aeronautics and Astronautics, Nanjin 210016, China.
Neuroinformatics
|March 19, 2013
Summary
This study introduces a novel semi-supervised multimodal relevance vector regression (SM-RVR) method to predict cognitive scores for Alzheimer's disease (AD) patients using brain imaging and biomarkers. The method shows promising results in evaluating disease progression and pathological stage.
Area of Science:
- Neurology
- Medical Imaging Analysis
- Machine Learning for Healthcare
Background:
- Accurate cognitive score estimation is crucial for tracking neurological disease progression, such as Alzheimer's disease (AD).
- Existing methods often lack multimodal integration or robust handling of data heterogeneity, particularly in mild cognitive impairment (MCI) subjects.
- Multimodal biomarkers including MRI, FDG-PET, and CSF offer rich information for disease assessment.
Purpose of the Study:
- To develop a novel semi-supervised multimodal relevance vector regression (SM-RVR) method for predicting clinical cognitive scores in neurological diseases.
- To leverage multimodal imaging and biological data for enhanced pathological staging and progression prediction in AD.
- To address the challenge of unstable clinical scores in MCI by utilizing their multimodal data in a semi-supervised learning framework.
Main Methods:
- A semi-supervised multimodal relevance vector regression (SM-RVR) approach was developed.
- The model utilizes multimodal data (MRI, FDG-PET, CSF) from Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
- A strategy for selecting informative MCI subjects was implemented to train the semi-supervised model.
Main Results:
- The SM-RVR method achieved a root-mean-square error (RMSE) of 1.91 and a correlation coefficient (CORR) of 0.80 for estimating MMSE scores.
- For ADAS-Cog scores, the method obtained an RMSE of 4.45 and a CORR of 0.78.
- The approach demonstrated promising performance in predicting cognitive scores for AD and normal control (NC) subjects.
Conclusions:
- The proposed SM-RVR method effectively predicts cognitive scores using multimodal biomarkers.
- The semi-supervised approach enhances prediction accuracy by leveraging MCI subject data.
- This method holds significant potential for evaluating disease stage and predicting progression in Alzheimer's disease studies.
