Dear Lynne,
there are several issues with your image that make the segmentation difficult or impossible:
1. The skull-stripped image contains a mix of NaN, negative values and values up to 50 in the background. If you skull-strip the image there should be only zeros in the background, otherwise the segmentation approach tries to also segment the background which will of course fail.
2. I would not recommend to use skull-stripped images for segmentation at all. Some gyri and the cerebellum are cut on your image and prior skull-tripping is not necessary for SPM12 segmentation and sometimes even cumbersome as in your case.
3. The segmentation approach is assuming 3 classes in the brain, but your image also contains a large lesion posterior to the left putamen which has the same intensity values as gray matter. Segmentation in the lesion will be therefore not correct.
After correcting the values in the background the segmentation looks much better, but you have to keep in mind that the deviating anatomy of your image (huge asymmetry in the ventricles, large lesion) will make a good segmentation quite difficult.
Best,
Christian
On Wed, 14 Jun 2017 16:09:57 -0400, Lynne Gauthier <[log in to unmask]> wrote:
>I have placed screenshots of the skull stripped and gray matter images as
>well as the preprocessed file in:
>
>https://drive.google.com/open?id=0B8ZAVitAc7uuSXFBMFFtMVZFR1U.
>
>Cross-hair is in the same place on both. This was best achieved
>segmentation with fwhm of 40 and bias correction of 0. As you can see, the
>GM values close to the edge of the brain are underestimated. With higher
>values of bias correction and/or higher fwhm, the GM values at the
>cross-hair were close to 0. fwhm of 30 yielded some odd classifications of
>wm in some of my images.
>
>Thanks!
>
>
>
>On Wed, Jun 14, 2017 at 4:03 AM, MRI More <[log in to unmask]> wrote:
>
>> Dear Lynne,
>>
>> Maybe you can upload one of the structural volumes somewhere so that we
>> can inspect the data / get an impression whether e. g. the tissue contrast
>> is unusual or not.
>>
>> Best regards
>>
>> Helmut
>>
>>
>
>
>--
>Lynne Gauthier, Ph.D.
>Assistant Professor of Physical Medicine and Rehabilitation
>The Ohio State University
>480 Medical Center Drive, 2145 Dodd Hall
>[log in to unmask]; [log in to unmask] <[log in to unmask]>
>Extension: 614-293-6287
>Rehabilitation Psychology Office: 614-293-3830
>Fax: 614-293-4870
>Cell: 857-998-7785
>Link to calendar:
>*https://www.google.com/calendar/embed?src=lynnevg%40gmail.com&ctz=America/New_York
><https://www.google.com/calendar/embed?src=lynnevg%40gmail.com&ctz=America/New_York>
>*(click week view for blocked schedule)
>*
><https://www.google.com/calendar/embed?src=lynnevg%40gmail.com&ctz=America/New_York>*
>
|