The HO atlas is already supplied as a “labeling” scheme in Lead-DBS so you could also use the single line:
>> ea_labeling2atlas(‘Harvard Oxford Thr25 2mm Whole Brain (Makris 2006)’)
to convert it into a subcortical atlas. This only works perfectly sometimes but it surely extracts all the labels as single files for you. When I just tried it here, I e.g. had to copy the file Left_Occipital_Pole.nii from the /midline subfolder to the /lh subfolder of the converted atlas set.
Alternatively, you can look at how other atlases are stored in Lead-DBS to find out how they should look like. Basically, each structure needs to be a single file and can be inside the /lh, /rh, /midline or /mixed subfolder. If e.g. the same homologue structure for lh and rh is stored in a single file, put it in mixed and lead-dbs will divide data at x=0mm.
Finally, it’s important to note that Lead-DBS uses the MNI 2009b NLIN 0.5×0.5×0.5mm space whereas FSL essentially doesn’t (it uses the older 6th gen nonlinear space).
Differences are small but can be significant in DBS imaging (read more about this issue here: http://www.lead-dbs.org/?p=1241).
Hope this helps,