I ran 3 identical first level FSL analyses of a localiser scan in order to compare the different bet settings (applied to all structural and functional images). I can confirm that BET with the recursive flag (-R) produces the best outcomes with higher Z values when transformed to standard space. The –B option looks good when examining the resulting BET images, and deals nicely with residual neck artefacts, but actually produces worse results at the level of the statistics.

Below are images showing zstats from each first level analysis (face-place localiser), converted to standard space and displayed on the MNI brain.

PPA (blue), FFA (orange), standard space – BET with no options

http://i.imgur.com/q3FdX.png

PPA (blue), FFA (orange), standard space – BET with –B option

http://i.imgur.com/QC9al.png

PPA (blue), FFA (orange), standard space – BET with recursive option (-R)

http://i.imgur.com/Nqob1.png

Activation is more extensive and more significant with the –R option.

Images of the structural scans after each bet extraction can be seen in the album below

http://laserdragon.imgur.com/brainimaging

It seems that the standard pipeline involves using the –R option anyway, but I thought people might be interested to see the difference it makes, and that the –B option turns out to be pretty bad.

Chris