Skip to main content
. 2014 Mar 14;8:24. doi: 10.3389/fninf.2014.00024

Figure 11.

Figure 11

Processing times for BROCCOLI for first level analysis using a permutation based t-test with 10,000 permutations (SPM, FSL, and AFNI do not provide any functions for first level permutation based analysis). Left: Voxel-level inference, the maximum t-test value is saved in each permutation. Right: Cluster-level inference, the extent of the largest cluster is saved in each permutation. A t-value of 3 was used as a cluster defining threshold. The data used is of the size 64 × 64 × 33 × 160. A brain mask was used to only perform the statistical calculations for the brain voxels. Note that these processing times do not include smoothing in each permutation. Smoothing the fMRI data 10,000 times takes about 8970 s using one core on the Intel CPU, 2710 s using all the four cores on the Intel CPU, 335 s with the Nvidia GPU and 550 s with the AMD GPU. Also note that ordinary least squares is used to estimate the GLM beta weights in each permutation, and not the more demanding Cochrane-Orcutt procedure.