Related Experiment Video
Updated: May 9, 2026
07:12
Using Informational Connectivity to Measure the Synchronous Emergence of fMRI Multi-voxel Information Across Time
Published on: July 1, 2014
The statistical analysis of multi-voxel patterns in functional imaging
Kai Schreiber1, Bart Krekelberg
1Center for Molecular and Behavioral Neuroscience, Rutgers University, Newark, New Jersey, USA.
Plos One
|July 18, 2013
Summary
Multi-voxel pattern analysis (MVPA) in BOLD imaging faces statistical challenges. A permutation test is crucial for accurate significance assessment in MRI experiments, avoiding inflated performance estimates.
Area of Science:
- Neuroimaging
- Cognitive Neuroscience
- Machine Learning in Neuroscience
Background:
- Multi-voxel pattern analysis (MVPA) is a growing technique in BOLD imaging to detect changes in activation patterns.
- The BOLD response's slow nature can lead to overestimated MVPA performance.
- Standard statistical tests are often invalid for typical MRI data.
Purpose of the Study:
- To address statistical challenges in MVPA for BOLD imaging.
- To present methods for avoiding overestimation of MVPA performance.
- To identify valid statistical tests for assessing significance in MVPA.
Main Methods:
- Tutorial presentation of methods to correct for BOLD response artifacts.
- Evaluation of standard statistical tests (e.g., Student's T, binomial) for validity.
- Implementation and assessment of a permutation test for statistical significance.
- Simulations to investigate the impact of temporal and spatial signal correlations.
Main Results:
- Standard statistical tests are invalid for typical MRI experiments.
- A carefully constructed permutation test correctly assesses statistical significance.
- MVPA performance estimates increase with temporal and spatial signal correlations.
- Methods to avoid overestimation of performance are presented.
Conclusions:
- Accurate statistical assessment in MVPA requires specialized methods like permutation testing.
- Signal correlations significantly influence MVPA performance estimates.
- Comparisons of MVPA performance across different conditions, subjects, or preprocessing methods demand careful consideration.
