Hi,
As these results are within 0.02% (not 2%, but 0.02%) of each other
then that is far less than the accuracy to which these methods work.
There was a slight numerical rounding error in the old version of
avwstats (no ++) when accumulating over large images. The newer
version of fslstats (which is almost identical to avwstats++ for these
features) should be more accurate. However, both are fine with
respect to numbers for siena and much less than the noise effects.
All the best,
Mark
On 2 Jan 2008, at 22:15, Bettyann Chodkowski wrote:
> Dear All,
>
> I think I found the source of the discrepancy. Currently I am
> using FSL v3.3.
>
> avwstats -v (no ++) returns a *slightly* different volume value
> than avwstats++
> fslstats (v4.0.1) returns the same volume value as avwstats++
>
> When I use avwstats (no ++), i get nearly identical numbers as to
> what is in the *.sienax file:
>
> % avwstats 20071129_13_1_stdmaskbrain_pve_0 -m -v
> 0.050212 7864320 6308232.703233
> % avwstats 20071129_13_1_stdmaskbrain_pve_1 -m -v
> 0.081955 7864320 6308232.703233
> % avwstats 20071129_13_1_stdmaskbrain_pve_2 -m -v
> 0.060953 7864320 6308232.703233
>
>>> csf = 0.050212 * 6308232.703233;
>>> grey = 0.081955 * 6308232.703233;
>>> white = 0.060953 * 6308232.703233;
>>> bpf = ( grey + white ) / ( grey + white + csf );
>>> fprintf( 'csf = %.1f grey = %.1f white = %.1f bpf=%.6f \n',
>>> csf, grey, white, bpf );
> csf = 316749.0 grey = 516991.2 white = 384505.7 bpf=0.739996
>
> From *.sienax file:
> Class: CSF tissue 1 tissue 2
> brain percentage
> Volumes: 316748.7 516990.9 384503.5
> 0.739996
>
> Results from avwstats++ are below, but bottom line:
> csf = 316824.7 grey = 516909.2 white = 384486.8 bpf = 0.7399
>
> I believe that the results generated from sienax/FSL v3.3 would be
> slightly different from
> sienax/FSL v4.0.1 due to the use of avwstats v fslstats, respectively.
>
> Is this something you see, too?
>
> Thanks,
> - BettyAnn
>
>
> On Wed, 2 Jan 2008 20:11:23 +0000, Bettyann Chodkowski
> <[log in to unmask]> wrote:
>
>> Mark,
>>
>> Ah, thanks. This helps a great deal.
>>
>> The values I calculate for the tissue classifications are still
>> not exactly the same as what I see
>> reported in the *.sienax file. The final brain percentage value
>> does match, however. Might you
>> see something I am doing wrong? i used fsl v3.3 when running
>> sienax but fsl v4.0.1 for fslstats
>> (which gives the same values as avwstats++ v3.3).
>>
>> % fslstats 20071129_13_1_stdmaskbrain_pve_0 -m -v
>> 0.050224 7864320 6308232.500000
>> % fslstats 20071129_13_1_stdmaskbrain_pve_1 -m -v
>> 0.081942 7864320 6308232.500000
>> % fslstats 20071129_13_1_stdmaskbrain_pve_2 -m -v
>> 0.060950 7864320 6308232.500000
>>
>> in matlab:
>>>> csf = 0.050224 * 6308232.500000;
>>>> grey = 0.081942 * 6308232.500000;
>>>> white = 0.060950 * 6308232.500000;
>>>> bpf = ( grey + white ) / ( grey + white + csf )
>> bpf = 0.7399
>>>> fprintf( 'csf = %.1f grey = %.1f white = %.1f \n', csf, grey,
>>>> white );
>> csf = 316824.7 grey = 516909.2 white = 384486.8
>>
>> The tissue volumes in my *.sienax file are:
>>
>> Class: CSF tissue 1 tissue
>> 2 brain percentage
>> Volumes: 316748.7 516990.9 384503.5
>> 0.739996
>>
>> Also, in the snippet of code from the sienax script where the
>> volumes are computed, are you
>> dividing the result by 1 in the dc calculator? if so, why?
>>
>> Thank you for your help,
>> - BettyAnn
>> ===========================================================
> =============
>
|