Hi Bruce

Thanks so much for your quick reply.

I had overlaid the ?h.defect_labels on the inflated.nofix and orig.nofix. I adjusted the threshold to focus on the largest defect. Although I could look at the defect, I couldn't figure how to fix it manually because the defect is too large.

In addition, I separately overlaid the ?h.defect_labels on the inflated.nofix and orig.nofix but I actually don't know how to use inflated.nofix to look for the defect. I don't know the difference between defect_labels overlaid on the inflated.nofix and orig.nofix.

Excuse me, I also want to ask another question. Because I also check the aseg.auto and aseg.presurf and find there is bad segmentation, can I only modify segmentation results manually or is there another approach to solve the bad segmentation problem? In some cases, the contrast of the image is not obvious and thus modifying segmentation results manually is hard.

I have put these two subject directories named Stroke_subject001.tar.gz and Stroke_subject001.tar.gz on your ftp.

Thank you so much for your help.

Best regards,

Jiun-Wei

On Mon, 9 Sep 2019 at 20:39, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote:

Hi Jiun-Wei

have you looked at the ?h.defect_labels overlaid on the inflated.nofix
and orig.nofix surfaces? That will tell you where the defect is that is
causing you trouble. If you have lesions they may be the source of the
issue. If you can't sort it out you can tar and gzip the entire subject
directory and put it on our ftp site for us to have a look
cheers
Bruce
On Mon, 9 Sep
2019, Jiun Wei Chen wrote:

>
> External Email - Use Caution
>
> Dear freesurfer experts,
> When executing recon-all in two subjects with brain lesions, the recon-all
> process stopped at CORRECTING DEFECT 13 (vertices=49319, convex hull=12973,
> v0=65403) and
> CORRECTING DEFECT 6 (vertices=23529, convex hull=6740, v0=16029),
> respectively.
>
> After reading the relevant problem from freesufer mailing list, I had
> checked the skull and cerebellum not attached to the surface, two
> hemispheres are separated using freeview. In addition, I also load
> the brain.mgz, wm.mgz, ?h.orig.nofix and ?h.inflated.nofix and separately
> overlap the ?h.defect_label onto ?h.orig.nofix , ?h.inflated.nofix to edit
> wm.mgz to solve this problem.
>
> However, after editing wm.mgz in two subjects, the recon-all process still
> stopped at CORRECTING DEFECT 0 (vertices=32853, convex hull=7926) and
> CORRECTING DEFECT 6 (vertices=16533, convex hull= 4510) , respectively.
>
> Although I try to erase the non-white matter and label voxels as white
> matter as much as possible, there are too many voxels I think I need to
> label, which may cause the failure.
>
> I can't figure out how to continue to solve this problem. I use freesurfer
> version 6.0.0.
>
> I really appreciate your help. Thanks a lot.
>
> Best regards,
>
> Jiun-Wei
>
>
>
>
>
>_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer