spm_hrf_boost.m

If the hrf doesn't fit properly the data, the magnitude of the beta value is biased. Some information can be recovered using the 1st and 2nd derivatives (http://www.frontiersin.org/Journal/10.3389/fnins.2014.00001/abstract). This function takes a SPM.mat and related files to create boosted beta parameters and con images. Those new images can then be entered into the 2nd level analysis. This usually improves your results by having more signal. Note that we constraint which voxel get boosted using the shift parameter. The default shift is 2.5 sec, which means that given the model that you specified for the hrf, +/- 2.5 at the peak is allowed. For instance a standard hrf peaks at 5 sec, then looking at the 1st derivative we can estimate when the hrf peaks - the signal will be boosted for those voxels with estimated peaks between 2.5 sec and 7.5 sec. This makes sense to apply this constraint (thanks to Donald McLaren for reminding me) as the 1st derivative can also fit some artefacts; at least in this range it is more probable to be signal.

FORMAT
spm_hrf_boost
spm_hrf_boost(SPM_location,shift)

INPUT
SPM_location is the full name of the SPM.mat (like cyril/home/docs/matlab/my_super_expe/SPM.mat)
shift is time shift allowed around the hrf peak - default is 2.5 sec

if no input, the user is prompted to select a SPM.mat and enter the time delay allowed around the hrf peak.

OUTPUT created in /hrf_boost
boost_beta_XXXX.hrd and boost_beta_XXXX.img corresponding to the boosted hrf
boost_con_XXXX.hrd and boost_con_XXXX.img that combined hrf regressors