Yes, there are two strategies you can use. You can either make a regular 3D grid spanning the whole brain in which only grid locations in grey matter are considered in the source estimation. The alternative is to construct a triangulated surface description of the cortical sheet.
It requires that you have a segmentation of your anatomical MRI, which specifies the different tissue types. Such a segmentation can be made from the anatomical MRI using ft_volumesegment (which internally uses the SPM8 toolbox). The anatomical MRI and its segmentation have to be aligned in the same coordinate system as the one in which you want to perform the source reconstruction. That often means that the MRI should be expressed in head coordinates.
If you want to use the canonical MRI that is included in SPM, you can also use the accompanying segmentation into the different tissue types. You can download the anatomical 1mm MRI and the segmentations from the MNI brainweb site. You should use the “crisp” segmentation for creating the grey matter volume. If you have SPM2 on your Matlab path, the MINC file format will automatically be detected and supported by FieldTrip.
FieldTrip does not have its own methods for this, but relies on external software such as BrainVoyager or FreeSurfer. The minumum-norm estimate tutorial has more details on this. You could also use the MATLAB isosurface function on a segmented brain, although the resulting surface topology will probably not be completely correct.