Re: [Freesurfer] improving white/grey segmentation in recon -all for better volume estmation

Hi Tilak

here's a sagittal snapshot of one of your subjects. You can see a number of problems with it:

1. The cerebellum is very bright and has little contrast. 2. The occipital lobe has a similar effect. 3. Superior regions don't show any gray/white contrast

the frontal lobes look ok (which I think is what you sent), but the rest of the volume doesn't have enough gray/white contrast to reliably and consistently estimate the class distributions, which is why things failed.

Can you tell us a bit more about the acquisitions? If I had to, I would guess that you have RF penetration issues and that the flip angle in occipital cortex is very different than what it is frontally. What transmit coil are you using? What receive coil? What field strength? Any acceleration?

cheers Bruce

On Sat, 26 Mar 2016, Ramtilak Gattu wrote:

Hey Bruce,

Thanks for the quick reply.

I have uploaded the cases in

https://waynestateprod-my.sharepoint.com/personal/ao3080_wayne_edu/_layouts /15/guestaccess.aspx?guestaccesstoken=%2bkyhy0gR4k7Q3AechqG%2bjqb89ALB94xWl pwBQnNAhG8%3d&docid=0a72431d82b4d4b9b924cccbdfb7f34f8

The voxel dimensions are 1x1x1 mm.

Thanks

Regards

Tilak

---

From: Ramtilak Gattu Sent: Friday, March 25, 2016 9:59 PM To: freesurfer@nmr.mgh.harvard.edu Cc: tilakgattu@hotmail.com Subject: improving white/grey segmentation in recon -all for better volume estmation

Hello all,

We have run recon -all on  some healthy control subjects T1 images but we  found out that the gray-white segmentation on half our subjects is bad and it leads to increased white matter volumes and decreased gray matter volumes. I have included a snap shot showing the difference in segmentation from bad to good for 3 cases

https://waynestateprod-my.sharepoint.com/personal/ao3080_wayne_edu/_layouts /15/guestaccess.aspx?guestaccesstoken=ajPIRobFq548RyhRJWPQ9q70vAXUGIJ9h5WPD E1klUw%3d&docid=0dd4546a8004f46578b7a2177b192806a

. The parameter in the  sequence are TE=4.92 ms, TR=19 ms, Flip angle =20 degrees. and use a snapshot one slice to show the good CNR.  We are using Freesurfer version 5.3.

Thanks

Regards

Tilak.