We have very noisy/low SNR structural scans from a sample of children and adolescents. We are having trouble with skull stripping that is sufficient for surface-based analysis (e.g. Freesurfer).  A large portion of pial matter is included along the dorsal and posterior areas of the cortex for nearly every subject. We have tried multiple options in AFNI skull strip and FreeSurfer skull strip, with little success. 

We have recently tried BET. When adjusting the fractional intensity threshold to be larger (so more aggressive skull stripping, at 0.6, 0.7, 0.8, ect.), it starts removing parts of the prefrontal cortex, but still leaves the pial matter.
--