[MINC-users] Classify

[Jon Erik Ween]

Andrew,

So I guess what I need to do is review the intensity histograms of my  
scans and review the results before and after nu_correct so that voxel  
intensities wind up in narrower bins (with a trimodal distribution)?  
I'm not seeing how one plots the histogram output from mincstats  
(maybe you use another program or that?) nor how to tweak nu-correct  
to optimize the normalization? My pipeline is as follows;

1) an initial run with nu-correct on whole head
2) an initial run with mincbet for a rough mask
3) a cleaner run with nu_correct using the mask from step#2, though I  
don't see any parameters to set to optimize this step.
4) a second mincbet step on the volume from step 3 for an optimal  
mask, subtract the non-masked areas from the original volume for step  
#5 (the brain from minbet leaves a lot o whitematter holes in  
infarcted brain, so mincmath -sub gives better results.)
5) A third nu_correct run on the skull-stripped brain from step #4

Thanks for any suggestions.

Jon

Soli Deo Gloria

Jon Erik Ween, MD, MS
Scientist, Kunin-Lunenfeld Applied Research Unit
Director, Stroke Clinic, Brain Health Clinic, Baycrest Centre
Assistant Professor, Dept. of Medicine, Div. of Neurology
     University of Toronto Faculty of Medicine

Kimel Family Building, 6th Floor, Room 644
Baycrest Centre
3560 Bathurst Street
Toronto, Ontario M6A 2E1
Canada

Phone: 416-785-2500 x3648
Fax: 416-785-2484
Email: jween at klaru-baycrest.on.ca


Confidential: This communication and any attachment(s) may contain  
confidential or privileged information and is intended solely for the  
address(es) or the entity representing the recipient(s). If you have  
received this information in error, you are hereby advised to destroy  
the document and any attachment(s), make no copies of same and inform  
the sender immediately of the error. Any unauthorized use or  
disclosure of this information is strictly prohibited.



On 5-Apr-09, at 9:47 AM, Andrew Janke wrote:

> Hi Jon,
>
>> 1) How to optimize the classification so the grey matter is more
>> representative of that seen visually on the 3DT1 (particularly the
>> cortical ribbon)?
>
> To get a better classification I think the first thing you are going
> to have to do is to remove more of the inhomogeneity in the GM/WM
> signal. classify expects the input data to have similar intensities
> for all the WM and GM voxels. Have these images been run through N3?
>
>> 3) Saving the tagfile in "Display" gives you a bit different format
>> (which classify doesn't like)  compared to the tagfile in ../mni/ 
>> share/
>> classify/ntags_1000_bg.tag
>>
>> Display "save labels .tag":  -82 -109 108 1 1 1
>> /mni/share/classify/ntags_1000_bg.tag:  -82 -109 108 "1"
>>
>> Am I using the wrong function here?
>
> No, I have always just mashed the files about to suit with a text
> editor. The tags from display are more about saving a list of 3D
> points rather than for use with classify.
>
>> 4) Once I get a classified volume (and I have flair lesion and 3DT1
>> lesion in separate masks) how do I calculate volumes? I can't seem to
>> find the functions that do this, though suspect they are in minmath  
>> or
>> minccalc somewhere.
>
> Close.
>
>   mincstats -count -mask_binvalue <x>
>
> Is probably what you want.
>
>
> --
> Andrew Janke
> (a.janke at gmail.com || http://a.janke.googlepages.com/)
> Canberra->Australia    +61 (402) 700 883
>
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www2.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>