[Brains-users] Too few grey matter plugs found

[Ian Malone]

Hi,

I'll see what I can do about datasets, though don't think I can give 
away the ones I've looked at so far.  In terms of numerical values, I've 
loaded the re-sampled volume up in ImageJ and they should be closer to 
55,75,24.  I'm not sure how to get at the brain mask being used, but 
having created an roi from a mask created from my defined Talairach 
parameters and the 'brain' Talairach box, it looks to cover most of the 
re-sampled brain (not right out to the surface) and produces the 
following measurements:

 2002_Unknown.nii.gz for brain-Mask.zroi traced by malone gave:
        Mean: 64.163005
        Mean-Absolute-Deviation: 13.058092
        Variance: 258.197096
        Standard-Deviation: 16.068513
        Skewness: 1.537886
        Kurtosis: -0.251059
        Minimum: 3.000000
        Maximum: 102.000000
        Min-Loc-X: 128
        Min-Loc-Y: 124
        Min-Loc-Z: 120
        Max-Loc-X: 111
        Max-Loc-Y: 117
        Max-Loc-Z: 84
        Center-Of-Mass-X: 127.352565
        Center-Of-Mass-Y: 141.574123
        Center-Of-Mass-Z: 83.716772

The 'patid' string is originating from the directory name, I've run 
through again using a positive integer instead, but with no change.  
I've also just noticed that using grey and white matter traces makes no 
difference to those initial estimates (the "b2 generate class-plugs" 
line), should it?

Thanks for the suggestions,
Ian

Greg Harris wrote:
> Hi, Ian.
>
> Is there any way you could donate one of the T1-only datasets that is
> causing this failure?
>
> The text {{80.186111 113.979065 30.087282 10.8}} gives the expected gray,
> white, and csf means for your T1 image, and the expected plug variance.
> The first failure of kmeans clustering seems to indicate that your brain
> image is not near these numerical estimates. When it continues past the text
>   
>> Class 1:  32.042
>> Class 2:  32.042
>> Class 3:  32.042
>>     
> This says to me that the mean, max and min in the brain mask (m1) were
> identical, so I would also want to examine the brain mask:  why is it all
> one intensity in the mask? The text
>   
>> Diagnostic: enclosing box is 69..186 x 82..192 x 15..176
>>     
> gives the bounding box for this mask, so it is not one point or something
> like that.  
>
> The text "r= patid" might be causing a problem:  r is the random seed for
> the plug picker.  Give it a positive integer, not a variable name.  We use
> our patient id solely to make the choice of training class plugs random but
> reproducible.
>
> Greg Harris
> Iowa Psychiatric Neuroimaging Consortium
>
> On 8/18/08 11:57 AM, "Ian Malone" <ian.malone at drc.ion.ucl.ac.uk> wrote:
>
>   
>> Hi,
>>
>> I'm trying to follow through the Brains2 standard workup, with T1 scans
>> only.  I keep running into the following problem when running tissue
>> classification:
>> "Failure occured (-1): Too few grey matter plugs found,
>> {CubicCentimeters 0.000000000000000} CC"
>>
>> I've tried several different T1 with the same result, using both the MR5
>> and MR6 settings for the reslicing. The only time I've successfully run
>> tissue classification was with the MNI 152 brain.  Other scans fail,
>> both with venous blood traces only and with basal grey and white matter
>> traces.  The above error will follow a series of steps reporting k-means
>> errors as follows:
>>
>> Remark:  min gm plugs == 4000
>> Remark:  min wm plugs == 2000
>> Remark:  min csf plugs == 200
>> b2 generate class-plugs i1 {gm wm csf} {{80.186111 113.979065 30.087282
>> 10.8}} m1 r= patid n= 4000 c= 0.85 p= 0.10 mo= 2.5 vo= 10.0 ps= 2.0 dx=
>> 1 dy= 1 dz= 1 gm= 4000 wm= 2000 csf= 200 pb= m4
>> Diagnostic:  bk_generateClassPlugs entered.
>> Diagnostic:  pv_plugCluster entered.  numClasses == 3
>> Diagnostic: plug dimensions  = 2 x 2 x 2,  plug volume = 8.000 voxels
>> Diagnostic: Random Number generator seed = 1769234800
>> Diagnostic: enclosing box is 69..186 x 82..192 x 15..176
>> Diagnostic: brain mask has 845974 voxels
>> Diagnostic: Image set plug variance table:  0 (10.800)
>> Diagnostic: chunk_z 15..176
>> Diagnostic: chunk_y 82..192
>> Diagnostic: chunk_x 69..186
>> Diagnostic: initial count1 = 849090 ,  oldcount1 = 845974
>> Diagnostic: max_points = 119999 ,  numplugs = 4014
>> Diagnostic: count1 = 849090,  count = 0,  sumplugs = 4014
>> Diagnostic: desired =  4014,  pick_from = 211,  band = 5
>> Diagnostic: desired =  1819,  pick_from = 211,  band = 10
>> Diagnostic: HARVESTED 5648 plugs in chunk 15..176 x 82..192 x 69..186.
>> Table of initial centers:
>> Class 1:  80.186
>> Class 2:  113.979
>> Class 3:  30.087
>> Diagnostic: ipl_KMNS produced IFAULT = 1
>> Table of initial centers:
>> Class 1:  32.042
>> Class 2:  32.042
>> Class 3:  32.042
>> Diagnostic: ipl_KMNS produced IFAULT = 1
>> Warning Diagnostic:  =====>  map of class mean order statistics failed
>> to resolve the desired number of classes!
>> Diagnostic: Center Table
>> Diagnostic: Center Table
>> Diagnostic: Center Table
>>
>>
>> Can anyone suggest what might be going wrong?
>>
>> Thanks for your time,
>> Ian
>>     
>
>
>