Related Experiment Video
Updated: Jun 12, 2026
07:12
Using Informational Connectivity to Measure the Synchronous Emergence of fMRI Multi-voxel Information Across Time
Published on: July 1, 2014
A mutual information-based metric for evaluation of fMRI data-processing approaches.
Babak Afshin-Pour1, Hamid Soltanian-Zadeh, Gholam-Ali Hossein-Zadeh
1Control and Intelligent Processing Center of Excellence, School of Electrical and Computer Engineering, University College of Engineering, University of Tehran, Tehran, Iran.
Human Brain Mapping
|June 10, 2010
Summary
We introduce a new mutual information (MI) metric to evaluate fMRI activation detection. This method is more sensitive than existing metrics on real datasets, improving analysis of brain activity.
Area of Science:
- Neuroimaging
- Functional Magnetic Resonance Imaging (fMRI)
- Statistical Analysis
Background:
- Evaluating activation detection performance in fMRI is crucial for accurate brain activity mapping.
- Existing metrics may lack sensitivity or generalizability in real experimental settings.
- Mutual Information (MI) offers a novel way to quantify information preserved between fMRI time-series and activation maps.
Purpose of the Study:
- To propose and validate a novel mutual information (MI)-based metric for evaluating fMRI activation detection performance.
- To compare the sensitivity of the proposed MI metric against existing performance metrics using simulated and real fMRI datasets.
- To assess the impact of group size and analysis software (FSL4, SPM5) on activation detection using the MI metric.
Main Methods:
- Calculated approximate mutual information (MI) between fMRI time-series and activation maps derived from independent training datasets.
- Compared MI metric sensitivity with ROC curves, reproducibility, and Jaccard overlap metrics on simulated and real group fMRI data.
- Investigated the relationship between MI, false discovery rate (FDR), and the fraction of active voxels across different analysis models (GLM).
Main Results:
- Mutual Information (MI) values were higher for larger subject groups (15 vs. 10 subjects) in real fMRI group analyses.
- The MI metric demonstrated greater sensitivity than reproducibility and Jaccard overlap metrics for activation map evaluation.
- At a fixed false discovery rate (FDR), the General Linear Model (GLM) using FSL4 extracted more voxels and information compared to SPM5.
Conclusions:
- The proposed mutual information (MI)-based metric provides a sensitive and robust evaluation of fMRI activation detection performance.
- MI is a valuable tool for comparing different analysis methods and understanding information preservation in neuroimaging studies.
- Findings suggest FSL4 may offer advantages over SPM5 in extracting more relevant information under specific FDR constraints.
