[Brains-users] Problems thresholding dark images

[Ronald Pierson]

I think it should just create an image, but with all the masks that are
created in the previous steps you need to turn them off to see the
image.

It sounds like the range is awfully low, though, and spreading out 29
original intensity values to create a new image may end up with a pretty
bad looking image. The optimal solution is to have a scan sequence that
gets rid of the bright areas in the fat or arterial blood, but that
isn't much help at this point.

Eugene, you and I talked yesterday about going back to the original raw,
12-bit (or float?) image to see if there was more dynamic range within
the brain there, and the bright areas may have intensities off in the
thousands. That may be the best option to try at this point. When the
image is first resampled the entire dynamic range of the whole scan is
crunched into 8 bits, 0-255. So, the bright stuff in the scan ends up
taking the upper range, and the region where the brain is only gets the
bottom 10% or so of the dynamic range. You really lose significant
figures in this conversion. So, if we can start our thresholding process
with the image in its original data type we can preserve the
significance and intensity spread of that data.

I think a process that would do we discussed is the following. Load the
raw image, the resample transform (T1_ACPC.xfrm), set the transform to
the image. Then save this image as a 16 bit or floating point image,
maybe with a temporary name (T1_temp.hdr). Do not use the Save Resampled
Image option, as that creates an 8-bit image only. Just use the Save
Image option, change the path to be in your 10_ACPC directory, and set
the data type to be 16 bit. In brains2 it should basically look the same
as your other T1, but the intensity values will be closer to the
original (you can check these using the Viewer => tracker). Exit and
restart brains2, load this new image, then do the steps to threshold it
(from the previous email) and save it with the proper name.


Ron

On Oct 07, 2005 07:32 AM, Eugene Zeien <eugene-zeien at uiowa.edu> wrote:

> Ron? correct me if I'm wrong, but I believe the "b2 clip image"
> creates
> both an image and a mask.
> In order to see the resulting image, you will need to go to the 
> "selector" (button at the bottom center
> of each image display pane), choose a mask, then click on the
> "display"
> button to hide that mask.
> This will need to be done for each mask.
> 
> Eugene Zeien
> 
> Kevin Berne wrote:
> 
> >Hello again,
> >Thank you for your insightful guide and for the time it took you to
> >write it. I have applied it to my images and found it working up to a
> >certain point, but not beyond. To be more specific, I loaded the
> >unresampled T1, applied my transform to it and also loaded the
> >talairach.bnd and .box. Converting the box to a mask and measuring
> >the image mask also worked without any problems, resulting in a
> >maximum intensity of about 29 (assuming I picked the right value out
> >of the many, it was fittingly named "maximum""). Adding the 20 you
> >mentioned to it would therefore bring us up to 49.
> >With the removal of bright stuff from the image the trouble basically
> >starts. After thresholding the image with 0, resulting in a uniform
> >whole-image mask and entering the requested line "b2 clip image i1 m2
> >upper-bound= $threshold" replacing $threshold with the value 49 into
> >the Advance Console, which our Brains2 does in fact contain, just
> >another uniform mask is created. Adding up the clipped images only
> >results in more uniform patterns. There seems no recognizable image
> >to be found after the application of the process, for whatever
> >reason. Measuring the resulting image displays minimum and maximum
> >values both as 49, 98 after adding it too itself.
> >If you would happen to know how this can be resolved, I would again
> >be very grateful, as always.
> >In the meantime, a nice weekend to everyone!
> >
> >Kevin Berne
> >ZI Mannheim
> >Central Institute of Mental Health
> >J 5
> >D-68159 Mannheim
> >Germany
> >
> >______________________________________________________________
> >Verschicken Sie romantische, coole und witzige Bilder per SMS!
> >Jetzt bei WEB.DE FreeMail: http://f.web.de/?mc=021193
> >
> >_______________________________________________
> >Brains-users mailing list
> >Brains-users at www.psychiatry.uiowa.edu
> >http://www.psychiatry.uiowa.edu/mailman/listinfo/brains-users
> >
> >  
> >
> 
> _______________________________________________
> Brains-users mailing list
> Brains-users at www.psychiatry.uiowa.edu
> http://www.psychiatry.uiowa.edu/mailman/listinfo/brains-users
>