fmri_csthreshold

Description

FMRI_CSTHRESHOLD Extracts critical cluster sizes from a CS distribution. Given a cluster-size (CS) null distribution (from fmri_csdist), extracts the critical cluster size at one or more family-wise error rate (FWER) levels. This is the third and final step of the Ledberg et al. (1998) procedure.

Usage

[CSIZE, report] = fmri_csthreshold(D, si_alpha, cs_alpha)

Inputs

D: CS distribution cell array {1 x R}, as returned by fmri_csdist.
SI_ALPHA: Statistical image alpha levels used to generate D. Default: [0.01 0.001 0.0001]
CS_ALPHA: FWER alpha level(s) for cluster-size thresholding. Default: [0.05 0.01 0.001]

Outputs

CSIZE: Cell array {1 x R}. Each cell is a (length(SI_ALPHA) x length(CS_ALPHA)) matrix of critical cluster sizes.
REPORT: Formatted string summarising the critical cluster sizes, suitable for printing (disp(report)) or saving to a file.
CS_table: Cell array {1 x R}, one table per regressor. Each table has rows = SI alpha levels and columns = FWER alpha levels. Columns are named like “CS_FWER_0_05”, “CS_FWER_0_01”, etc., and an extra column “SI_alpha” shows the statistical image thresholds.

Examples

CS_table{1} is a table for regressor #1.
% Full pipeline
acf   = fmri_acfestimate(datapath, regressor, 10);
D     = fmri_csdist(acf, [0.001 0.0001], 200);
[CS, report] = fmri_csthreshold(D, [0.001 0.0001], [0.05 0.01]);
disp(report)
% Apply threshold to a z-map
z_thresh = zmap > norminv(1 - 0.001);
clean    = fmri_cleanclusters(z_thresh, CS{1}(1,1), 0);

References

Ledberg, A., Akerman, S., & Roland, P.E. (1998). NeuroImage, 8, 113-128.