Evaluating Approaches for Inference Testing of Whole-Brain Densely Sampled Single-Subject Task fMRI Data

Area of Science:

• Neuroimaging
• Systems Neuroscience
• Functional Magnetic Resonance Imaging (fMRI)

Background:

• Task-based fMRI enables precise mapping of individual brain activity patterns.
• Current methods often focus on cortical activation, limiting systems-level understanding.
• Statistical analysis approaches for densely sampled, whole-brain fMRI data require further development.

Purpose of the Study:

• To evaluate statistical analysis methods for whole-brain, task-based fMRI data.
• To compare the sensitivity and stability of different inference approaches for individual functional localization.
• To investigate the auditory system, including subcortical components, using multi-echo, densely sampled data.

Main Methods:

• Collected whole-brain, multi-echo, densely sampled fMRI data for the auditory system.
• Evaluated whole-brain approaches: voxel-level and cluster-level inference with varying thresholds, non-parametric permutation inference.
• Assessed region-specific approaches: top % t-statistics and non-parametric permutation inference.

Main Results:

• Whole-brain voxel-level analysis with false discovery rate (FDR) correction (p<0.05) demonstrated the highest sensitivity and consistency in detecting auditory regions across subjects and scan durations.
• Region-specific top % t-statistic analysis showed utility as an exploratory tool.
• Both whole-brain and region-specific approaches provided valuable insights into functional localization.

Conclusions:

• Whole-brain voxel-level analysis with FDR correction is a robust method for individual functional localization in task-based fMRI.
• Region-specific top % t-statistic analysis can complement standard methods for exploratory functional mapping.
• Further research into statistical analysis is crucial for advancing systems-level fMRI studies.