Precision fMRI and cluster-failure in the individual brain.

Area of Science:

• Neuroimaging
• Functional Magnetic Resonance Imaging (fMRI)
• Signal Processing

Background:

• Advances in neuroimaging, including higher magnetic field strengths and improved denoising, enhance functional MRI (fMRI) resolution.
• Ultrahigh spatial and temporal resolution in fMRI can reduce the temporal signal-to-noise ratio (tSNR), impacting sensitivity to blood oxygen level-dependent (BOLD) signals.
• tSNR variability across the brain necessitates robust methods for accurate functional contrast analysis in individuals.

Purpose of the Study:

• To investigate smoothing filters for enhancing BOLD sensitivity in fMRI.
• To assess the preservation of spatial accuracy in activated clusters during single-subject fMRI analysis.
• To introduce and validate a novel smoothing technique, adaptive-weight smoothing with optimized metrics (AWSOM), for precision fMRI.

Main Methods:

• Evaluation of various smoothing filters for BOLD sensitivity and spatial accuracy.
• Development and application of adaptive-weight smoothing with optimized metrics (AWSOM) using a local inference approach.
• Comparison of AWSOM with traditional cluster-corrected inference on fMRI data with varying tSNR levels.

Main Results:

• Cluster correction in single subjects showed inflated family-wise error and false positive rates.
• AWSOM demonstrated comparable sensitivity to large-kernel Gaussian smoothing but preserved finer spatial details across different tSNR levels.
• AWSOM effectively suppressed false positives, maintained sensitivity to small activation clusters, and preserved the relative activation strength of significant voxels.

Conclusions:

• AWSOM offers a robust solution for improving BOLD sensitivity and spatial accuracy in precision fMRI.
• The method is effective in mitigating false positives and preserving signal integrity, crucial for whole-brain analyses with heterogeneous tSNR.
• AWSOM is a valuable tool for diverse fMRI applications, now freely available to the research community.