BrainVISA

I have nice Siemens Trio-MPRAGE data of more than 10 brains scanned three times each.
Although realignment & averaging of three acquisitions yields a high quality image, unfortunately, they seem not optimal for surface reconstruction.
(Those data were originally acquired for a VBM study in the rostral part of the brain.)

The main problem is that Vip Get Brain process includes the dura mater in the ccipital lobe
possibly because its intensity range is almost the same as the white matter (The histgram seems OK to me).
Consequently the hemi_cortex show erosion into the occipital dura.

I found some improvement by doing VipBiasCorrection for each of the three MPRAGE scans BEFORE realignment & averaging.
I also tried some filtering and deblurring of the images. But...

Are there any workaround other than manual editing or using volume coil?
I suspect the occipital dura has a uniform intensty enough to be eliminated automatically in good part.

Well,
I have my usual joker, everything should beb etter in the next release of the pipeline, but we do not know when...

Otherwise, my own experience lead be to be very careful relative to averaging images: when the registration is not perfect or because of variations of partial volume, the result look nicer but some properties are lost. For instance, a one voxel thick layer of black voxels between brain and dura may become grey. Therefore, I advise you to process only one dataset, you may be more successful.

Otherwise, you can post one of your dataset somewhere on the web for me to have a look.

Thank you for the helpful suggestion. Maybe I expected too much because averaging works well on our volume coil data and the resulting mesh is more than excellent with BrainVISA!

That is not dura you're seeing - but CSF flow. For you future image aquvisition just vary your TI over a small range - that will take away the TOF hyperintensity effect.