[Freesurfer] Segmentation problem

Hi,

I have a brain segmentation problem.

Scanner: Siemens 3T Verio Sequence: MPrage, TR=1900ms, TE=2.45ms, TI=900ms, Voxel: 1x1x1mm, 192 sagittal slices.

Problem: The lay of dura mater between brain and skull is bright (See attached brain.jpg). As a result, the segmented brain.mgz has a "cap", and the pial surface and GM/WM surface will "leak" to the dura layer somewhere (see the red dots in surfaces.jpg). This happened to some places and needs a lot of manual work to remove the connection between the brain and dura.

I tried to increase the TR to 3000ms. Still some connection somewhere, although the results are much better. Anybody optimized the mprage sequence to minimize the problem?

Thank you.

XJ Kang