fmri_tfce

Description

FMRI_TFCE Threshold-free cluster enhancement (TFCE) of a statistical map. Computes the TFCE score for each voxel in a statistical map. TFCE integrates cluster extent and peak height across all possible thresholds, avoiding the need to choose a fixed cluster-forming threshold. Higher TFCE scores reflect voxels that are both part of large clusters and have high signal intensity. The TFCE statistic is defined as: TFCE(v) = integral[ e(h)^E * h^H dh ] where e(h) is the cluster extent at threshold h, E weights cluster extent (default 0.5), and H weights peak height (default 2).

Usage

tfce = fmri_tfce(data)
tfce = fmri_tfce(data, E, H, N)

Inputs

DATA: Statistical map, vectorized (228453 x 1). Must be non-negative; run separately for positive and negative tails if needed.
E: Extent exponent. Default: 0.5 (Smith & Nichols recommendation).
H: Height exponent. Default: 2 (Smith & Nichols recommendation).
N: Number of threshold bins for numerical integration. Default: 50. More bins = more accurate but slower.

Outputs

TFCE: TFCE map, vectorized (228453 x 1)

Examples

% Positive tail
tfce_pos = fmri_tfce(max(zmap, 0));
% Negative tail
tfce_neg = fmri_tfce(max(-zmap, 0));
% Visualize
fmri_showslices(tfce_pos, 1, 2, [0 max(tfce_pos)])

References

Smith, S.M. & Nichols, T.E. (2009). Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. NeuroImage, 44(1), 83-98.