Task-correlated facial and head movements in classifier-based real-time FMRI

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Real-time fMRI is susceptible to movement artifacts, as standard artifact removal techniques are not applicable.
• Multi-voxel classifier methods in real-time fMRI are particularly sensitive to these artifacts.
• Understanding movement-induced artifacts is crucial for accurate real-time fMRI interpretation.

Purpose of the Study:

• To systematically investigate how head and face movements affect multi-voxel classifiers in real-time fMRI.
• To quantify the extent to which movements can be misinterpreted as neural signals.
• To assess the impact of movement artifacts on reported classifier accuracy in real-time fMRI.

Main Methods:

• Ten subjects performed twelve instructed movements during fMRI scans.
• Retrospective analysis of data from a prior real-time fMRI study was conducted.
• Whole-brain classifiers and spatial activation maps were analyzed for movement-related artifacts.

Main Results:

• Classifiers based solely on movements produced false positives in all tested cases (P < .05).
• Spatial activation maps showed artifacts for two of the twelve movement tasks.
• While neural activity mostly explained high accuracies, some results were likely dominated by movement artifacts.

Conclusions:

• Various movements, including eye, face, and body motions, can generate false positives in real-time fMRI.
• Movement artifacts pose a significant challenge to the reliability of classifier-based real-time fMRI.
• Careful consideration of movement artifacts is essential for valid real-time fMRI research.