Single-trial variability in event-related BOLD signals

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Analysis

Background:

• Current fMRI analysis methods rely on assumed, uniform hemodynamic responses (HRs) and noise patterns.
• These assumptions fail when HRs vary unpredictably across brain regions or experimental trials.
• An alternative approach is needed to capture dynamic, trial-specific brain activity.

Purpose of the Study:

• To employ Infomax independent component analysis (ICA) for detecting and visualizing single-trial hemodynamic response (HR) variations in event-related fMRI data.
• To challenge the homogeneity assumptions common in standard fMRI analysis.

Main Methods:

• Utilized Infomax ICA on blood oxygenation level-dependent (BOLD) fMRI data from six subjects undergoing visual stimulation (flickering checkerboard).
• Stimuli varied in duration (0.5s short, 3s long) presented at 30s intervals.
• Developed novel BOLD-image plots to visualize component HRs across trials and sessions.

Main Results:

• ICA identified independent components in primary visual cortex (V1) and medial temporal cortex (MT/V5).
• BOLD-image plots revealed substantial, systematic variations in component HRs across trials, sessions, subjects, and brain areas.
• Unexpectedly, V1 component HRs showed two positive peaks for short stimuli in 4/6 subjects, contrasting with single peaks for long stimuli.

Conclusions:

• ICA combined with BOLD-image visualization effectively reveals significant, unforeseen hemodynamic response variations.
• This approach offers a powerful alternative to traditional methods relying on fixed HR templates and averaging.
• Findings highlight the dynamic nature of neural responses and the limitations of assuming response homogeneity.