Hi Charles

where do you landmark? From the image you sent the inferior part of the cerebellum is *very* dark. This is typical of falloff if the cerebellum is far from isocenter (for example if you landmark on the forehead ). If you really want cerebellum you probably need to landmark lower.

Also note that interpolating to 512 doesn't buy you anything - you are just upsampling the same data. It's not really .45m pixels

On Thu, 1 Aug 2013, charles laidi wrote:

Dear Freesurfers,

I am experiencing recurrent issue while performing cerebellar segmentation on T1 images.

These are the characteristics of the acquisition:

§  T1: field of view (FOV) = 230 mm

§  repetition time (TR) = 1140 msec

§  echo time (TE) = 4.38 msec

§  matrix size 256 × 256

§  interpolated to 512 × 512

§  yielding an in-plane voxel size of 0.45 mm × 0.45 mm

§  slice thickness of 0.9 mm

After recon-all (ended without errors) I always end up with the same error in segmentation (see attached file) : there is the same defect of segmentation in the inferior part of cerebellum.

Are you aware of any way to correct this defect ? (I think this might be related to the low resolution of my images)

Thank you for your time.

Best regards

Charles