I'm just shooting from the hip here, but here is what I would do: 1. Create a mask of writing mri_binarize --i 001.mgz --min 4000 --o mask.mgz mri_binarize --i 001.mgz --min 4000 --inv --o mask.inv.mgz 2. Mask out the writing mri_mask 001.mgz mask.inv.mgz 001.masked.mgz 3. Scale the mask to be something near gray/white matter fscalc mask.mgz mul VAL mask.scaled.mgz 4. Create a new image by adding the masked with the scaled mask fscalc 001.masked.mgz add mask.scaled.mgz -o 001.unmasked.mgz 5. Smooth that mri_fwhm --i 001.unmasked.mgz --fwhm 5 --o 001.unmasked.sm05.mgz --smooth-only 6. Mask this to only have writing voxels mri_mask 001.unmasked.sm05.mgz mask.mgz001.unmasked.sm05.masked.mgz 7. Create the final image by adding that to the masked fscalc 001.masked.mgz add 001.unmasked.sm05.masked.mgz -o final.mgz

You may have to play around with the amount of smoothing, I was just guessing at 5 good luck doug

On 08/22/2012 02:31 PM, Alexandra Tanner wrote:

Hi Doug,

Thanks for the quick reply! It seems the value of the writing is much higher than any value elsewhere in the image (over 4000). Instructions on how to replace this value with values of nearby voxels would be great -- hopefully that will allow us to run the autorecon1 command on it.

Thanks, Alex

...

Hi Alex, there's no way to "erase" it in a way that you can recover the underlying signal. I'm surprised that such small things are causing the talairach to crash. If the writing is always the same value and that value does not appear in other parts of the image, you might be able to replace the value with values of nearby voxels. Let me know if you need some instructions on this. doug

On 08/22/2012 02:08 PM, Alexandra Tanner wrote:

...

Hi Doug,

I'm currently working with 4-6 year old clinical pediatric non-Siemens DICOM files that I'm unpacking and preprocessing in freesurfer. So far we've been able to unpack the DICOMS to .nii format and successfully begin the autorecon process for our first few subjects. We've noticed, when looking at the raw scans for some of our subjects, however, there is writing (probably information from the scan console) included in the images, and in some slices, slightly overlapping with the outer edge of the skull/brain. When attempting to run the autorecon1 on these scans, the command terminates during the talairach transform. We're thinking, since we don't have this issue with all of our subjects, the writing may be what is causing the problem.

I'm wondering if you know of any way to remove the writing from the images (I tried seeing if I could erase the lettering in tkmedit but it seems it can't be erased manually). Attached is a screenshot of a coronal slice with the writing. Any suggestions would be greatly appreciated!

Thanks, Alex

-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html