Related Experiment Video
Updated: Jun 15, 2026
08:19
Simultaneous Data Collection of fMRI and fNIRS Measurements Using a Whole-Head Optode Array and Short-Distance Channels
Published on: October 20, 2023
Sparse logistic regression for whole-brain classification of fMRI data.
Srikanth Ryali1, Kaustubh Supekar, Daniel A Abrams
1Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA 94305, USA. sryali@stanford.edu
Neuroimage
|March 2, 2010
Summary
A new method using logistic regression with L1 and L2 regularization accurately identifies brain regions for fMRI data analysis. This approach improves discrimination between cognitive conditions and is computationally efficient.
Area of Science:
- Neuroimaging
- Machine Learning
- Cognitive Neuroscience
Background:
- Multivariate pattern recognition is used for fMRI analysis to detect brain activity patterns differentiating cognitive states.
- High dimensionality in fMRI data (many regions vs. few observations) limits method performance.
- Existing methods struggle with overfitting or computational complexity.
Purpose of the Study:
- To introduce a novel logistic regression method combining L1 and L2 regularization for fMRI data analysis.
- To improve the accurate estimation of discriminative brain regions across conditions.
- To address the dimensionality challenge in fMRI analysis.
Main Methods:
- Developed a logistic regression model incorporating both L1 and L2 norm regularization.
- L1 norm ensures fast, sparse, and generalizable solutions.
- L2 norm accounts for correlated brain regions, crucial for fMRI data.
Main Results:
- The novel method outperformed existing techniques on simulated data across various signal-to-noise ratios.
- On experimental fMRI data, it effectively isolated a fronto-temporal network differentiating music and speech stimuli.
- Demonstrated superior performance compared to L1-only regularization and SVM-based feature elimination.
Conclusions:
- The proposed method is computationally efficient and effective for fMRI data analysis.
- It successfully identifies relevant discriminative brain regions and accurately classifies cognitive conditions.
- Highlights the importance of combined L1/L2 regularization for capturing distributed neural patterns.
