Hi,
I'm mainly interested in hippocampus. FIRST results are ok, but I thought
I could improve those results using the other contrasts I already have,
that's where FAST came in. I was wondering if mfast segmentation would
perform better than FIRST's since you can add more channels.
Thank you ,
Diana
Brian Patenaude dijo:
> Hi,
>
> FIRST and mfast both give the segmentations in the native image
> space. FIRST applies the inverse im1_to_std_sub.mat to the shape
> models to transform the models into native space.
> By "mask that to the mfast output", can you be a little more precise.
> Also, what structures are you looking at ? Have you observed and area
> where FIRST is not performing well, such that it may require the mask?
>
> Cheers,
>
> Brian
>
> On 19 Feb 2008, at 11:55, Diana Tordesillas Gutiérrez wrote:
>
>> Brian,
>>
>> Thank you for the fast reply.
>>
>> What do you think of the following pipeline:
>>
>> - $first_flirt on T1
>> - $run_first on T1
>> -Apply the inverse im1_to_std_sub.mat to the output_image and mask
>> that to
>> the mfast output. Then use fslmaths to extract volume
>> values....would that
>> give a better registration and segmentation or that's an incorrect
>> pipeline?
>>
>> Diana
>>
>>
>>
>> -------------------------------------------------
>>
>> Brian Patenaude dijo:
>>> Hi,
>>>
>>> There is no direct means of incorporating other modalities into
>>> FIRST. The training data
>>> for the model only had the T1-weighted images and thus first expects
>>> T1 contrast. One simple use of another contrast, could be to use it
>>> for the boundary correction (provided that it was less noisy and
>>> better contrast).
>>>
>>> Cheers,
>>>
>>> Brian
>>>
>>>
>>> On 18 Feb 2008, at 12:07, Diana Tordesillas wrote:
>>>
>>>> Dear FSL users,
>>>>
>>>> I’ve just learned successfully how to run FIRST on my subjects
>>>> using the T1
>>>> images, but since I also have T2 and PD images I wonder if I would
>>>> have
>>>> better results if I use all information I have.
>>>>
>>>> I would like to know if it’s possible to use the segmented image
>>>> (mfast
>>>> output) as im1 input in first_flirt and further analysis or there
>>>> is a
>>>> better way to introduce that information in FIRST.
>>>>
>>>> Thanks in advance,
>>>>
>>>> Diana
>>>>
>>>
>>
>
|