OK,thank you very much!!
发自我的 iPhone
在 2012-9-24,下午8:08,Rolf Heckemann <[log in to unmask]> 写道:
> Hi
>
> You could use label propagation to create a white-matter segmentation
> of your MPRAGE image that includes lesions, at least if those lesions
> are surrounded by normally intense white matter. I've done this using
> FSL and IRTK -- the procedure is described in some detail in Heckemann
> et al., Neuroimage 2011 (open access link:
> http://dx.doi.org/10.1016/j.neuroimage.2011.03.014 -- scan for WMHI)
> -- but here's a somewhat simplified outline using just FSL tools and
> no external atlases. N.B.: this may or may not work for you -- I
> haven't tried this exact approach.
>
> 1. You will need some MPRAGE images -- say nine -- of subjects without
> lesions, preferrably acquired on the same scanner. Let's call these
> a1..a9.
>
> 2. Segment these with fast -g aX
>
> 3. Register each MPRAGE with your target, using flirt and fnirt.
>
> 4. Apply the resulting transformations to each corresponding grey
> matter label (aX_seg_1)
>
> 5. Sum the transformed labels with fslmaths -add and threshold the
> result at 5 (9/2 rounded)
>
> 6. Optional: subtract the original FAST grey matter label from the
> output of 5. to obtain lesion labels. You may have to apply an
> erosion or opening (available in fslmaths) to remove scattered false
> positives.
>
> Please let me know how it goes.
>
> Rolf
>
>
> On 19 September 2012 06:10, Ai Edward <[log in to unmask]> wrote:
>> Hello
>>
>> I am using FAST to segment GM, WM in patients with white matter lesions(WMLs). Image type is MPRAGE. But I found FAST mistakes most of WMLs as GM, and the rest of WMLs as WM. How can I solved this problem?
>>
>> FSL 5.0 just released. Is there a good way to segment WMLs in new version?
>>
>> Thank you very much!
>>
>> Edwardai
|