Hi,
This is probably as good as it gets.
Cheers,
Brian
On 10 Jul 2008, at 00:38, Jay Ives wrote:
> Hi Brian,
>
> I presume one doesn't have to run the thalamus segmentation
> separately - that it is done by the -intref option.
>
> I ran this with 30 and 60 modes, but still there are several voxels
> outside the model. Is this as good as it gets?
>
> Thx...................J
>
> ----- Original Message ----- From: "Brian Patenaude"
> <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Wednesday, July 09, 2008 6:26 PM
> Subject: Re: [FSL] FIRST Hippocampus Segmentation
>
>
>> Hi,
>>
>>> OK. Does one specify the model with the -m flag in addition to
>>> the -intref option? In essence, what is correct the command line
>>> format when using -intref?
>>
>> -intref ${FSLDIR}/data/first/models_317_bin/L_Thal_bin.bmv -m $
>> {FSLDIR}/data/first/models_317_bin/intref_thal/L_Hipp_bin.bmv
>>
>>
>>
>>> What do you mean by "leave-one-out cross-validation "
>>
>> leave-one-out cross-validation is a fairly standard technique for
>> assessing accuracy in an nearly unbiased manner (particularly when
>> sample sizes are not large). The model is trained from all
>> subjects but one, then applied to the left out subject. This is
>> then repeated for each subject. By doing this, the test subject is
>> never included in the training data.
>>
>> Cheers,
>>
>> Brian
>
|