Related Experiment Video
Updated: Nov 26, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
16.0K
Bayesian variable selection in logistic regression with application to whole-brain functional connectivity analysis
Xuan Cao1, Kyoungjae Lee2, Qingling Huang3
1Division of Statistics and Data Science, Department of Mathematical Sciences, University of Cincinnati.
Statistical Methods in Medical Research
|December 14, 2020
Summary
This study introduces a new Bayesian model using functional MRI (fMRI) radiomics to predict Parkinson's disease. The advanced method achieved high accuracy, aiding in early diagnosis and identifying key brain regions.
Area of Science:
- Neuroimaging
- Biostatistics
- Radiomics
Background:
- Parkinson's disease (PD) is a neurodegenerative disorder diagnosed clinically and via MRI.
- Current diagnostic methods can be improved with advanced analytical techniques.
Purpose of the Study:
- To propose a Bayesian variable selection model for Parkinson's disease prediction using functional MRI (fMRI) radiomics.
- To evaluate the model's performance against existing methods and identify discriminative brain regions.
Main Methods:
- Developed a Bayesian logistic regression model with spike and slab priors for variable selection.
- Employed an approximate Gibbs sampler using a t-distribution to replace the logistic distribution.
- Extracted 6216 whole-brain functional connectivity features from fMRI data of 70 PD patients and 50 healthy controls.
Main Results:
- The proposed Bayesian model demonstrated superior performance with an average prediction accuracy of 0.83.
- Simulation studies confirmed the model's selection consistency and outperformance over state-of-the-art methods.
- Identified specific brain regions most discriminative for Parkinson's disease.
Conclusions:
- The Bayesian radiomics approach using fMRI shows significant potential for supporting the radiological diagnosis of Parkinson's disease.
- The method offers improved prediction accuracy and aids in understanding disease-related brain alterations.
