A Multi-Dataset Evaluation of Frame Censoring for Motion Correction in Task-Based fMRI

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Subject motion is a significant confound in functional magnetic resonance imaging (fMRI), impacting signal accuracy.
• Frame censoring, excluding motion-contaminated data, is increasingly used but its effectiveness requires systematic evaluation.
• Limited research exists on comparing frame censoring efficacy against other motion correction techniques in task-based fMRI.

Purpose of the Study:

• To systematically compare the performance of frame censoring with other common motion correction strategies for task-based fMRI.
• To evaluate these methods across diverse datasets, tasks, and participant age groups using reproducible workflows.

Main Methods:

• Analysis of eight publicly available fMRI datasets (11 tasks, child to adult participants).
• Comparison of frame censoring (using FD and DVARS thresholds) against 6/24 motion regressors, wavelet despiking, robust weighted least squares, and ICA-based denoising.
• Performance metrics included group analysis maximum t-values and single-subject region of interest activation/reliability.

Main Results:

• Modest frame censoring (1-2% data loss) showed consistent improvements over standard motion regressors.
• Performance gains from frame censoring were often comparable to other tested motion correction techniques.
• No single motion mitigation strategy demonstrated consistent superiority across all datasets and tasks.

Conclusions:

• Frame censoring is a viable strategy for improving task-based fMRI motion correction, particularly at low data loss percentages.
• The optimal motion correction method is context-dependent, influenced by the specific fMRI dataset and the chosen outcome metric.
• Researchers should carefully consider dataset characteristics and analysis goals when selecting a motion mitigation strategy.