Groupwise structural sparsity for discriminative voxels identification.

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Functional Magnetic Resonance Imaging (fMRI) is crucial for studying brain activity.
• Identifying discriminative voxels for learning and aversive conditioning is challenging due to data limitations.
• Existing methods risk false or missed detections in complex neuroimaging datasets.

Purpose of the Study:

• To develop a robust voxel selection method for fMRI data.
• To identify brain regions involved in human learning under aversive conditioning.
• To address challenges in acquiring sufficient sample sizes for psychological experiments.

Main Methods:

• A stable hierarchical voting (SHV) mechanism based on stability selection was proposed.
• The method evaluates spatial random sampling quality and minimizes detection errors.
• Performance was assessed using simulated and public fMRI datasets.

Main Results:

• The SHV algorithm successfully identified sparse and related patterns across subjects in fMRI data.
• Stable weight maps were generated for different phases of fear conditioning.
• Regularization strategy significantly impacted the interpretability of results.

Conclusions:

• The SHV mechanism provides a reliable approach for voxel selection in fMRI studies.
• Aversive conditioning causally alters visual-cortical activity, as evidenced by the identified patterns.
• The findings have implications for understanding learning and fear responses in the brain.