SCI Seg3D Mailing List

Text archives Help


[Seg3D] Re: Re: Re: Re: Re: Re: Re: Re: Re: Mask Data Filter


Chronological Thread 
  • From: Kristen Zygmunt <krismz@sci.utah.edu>
  • To: seg3d@sci.utah.edu
  • Subject: [Seg3D] Re: Re: Re: Re: Re: Re: Re: Re: Re: Mask Data Filter
  • Date: Mon, 21 Jun 2010 18:58:52 -0600
Hi Elizabeth,
I was able to look at the data and while the masked data appears to be binary, there are actually many different values in the masked data with values between 32644 and 33094. You can see this by masking the data, then selecting your original data volume. Go to Tools -> Threshold Tool and set lower to 32644 and upper to 33094. If you move the upper slider to a slightly lower value, you will start to see some white appear from the masked data, this will also happen if you move the lower slider to a slightly higher value. Another way to see that the masked data is not binary is to select the masked data and then run Filters -> Histogram Equalization. This will spread out the data values more so that they are represented by a wider variety of grayscale levels.

What I think is happening is either when loading your data into ImageJ or writing it out from ImageJ, ImageJ is quantizing the data to 16-bit unsigned values. When this happens, it spreads the data values out to the range 0 to 65535 and then Seg3D applies a color map to this data. Because the data you are interested in have values close together (32644-33094) compared to the 0-65535 range, the color map displays this data with one or possibly two very similar colors.

In other words, Seg3D uses 256 different colors to represent data. So in the range 0-65535, the color will change every time the data changes by 256, values 0-255 will have the same color, then values 256-511 will have the next color, etc. For the data that has not gone through the ImageJ process, the data range is in this case -2536 to 6043 which leads me to believe that the max possible range of your data in .img format is -8192 to 8192. This means the color will change every time the data changes by 64, values -8192 through -8129 will have the same color, then values -8128 through -8064 will have the next color and so on. Therefore, your example .img file uses some where around 8 different colors to represent the values in the masked region.

If you do the Filters -> Histogram Equalization for your .img example and your .tiff example, you will see that the masked area looks very similar. Depending on the processing or analysis you are performing after the masking, it may or may not matter that your data has different absolute ranges if coming from .img vs .tiff. If it does matter to you, I'd suggest either seeing if you can get ImageJ to output your data differently or sending all of your data through ImageJ. Unfortunately, Seg3D does not currently allow you to edit the color map to improve the display, though we are looking at what it would take to implement this feature.

Please let me know if you have any more questions,
Kris

On 6/21/10 11:55 AM, eschn1@umbc.edu wrote:
Hey Kris,
         I was able to log in anonymously and upload the two files into the
upload folder. The file names are Session with Mask Data Issue
(tiff file).ses and Session without MD issue (img file).ses .
However, when I tried to upload the files to the ForKris folder,
it would not allow it (even if I login again anonymously) saying I
wasn't authorized to access the server.  Anyways, the files should
be there in upload folder. Let me know if you need anything else.
Elizabeth


Hi Elizabeth,
      Maybe there's a problem with the directory I created.  Try ftping
to ftp.utah.sci.edu with username: anonymous and then cd into upload.
If possible, try cd into ForKris as well.  The server here should be
setup for anonymous ftp and has been tested recently.  If this still
doesn't work, perhaps we can setup a screen sharing session.
-Kris

On 6/21/10 8:39 AM, eschn1@umbc.edu wrote:
Hi Kris,

I wanna try ftping you two files (one session with an .img file that
works
and another session with .tiff that doesn't) to your folder but right
now
I do not have any login id for your ftp server so I am not authorized to
upload files. It also won't let me login anonymously. It there a way
that
I can gain authorization to your ftp server so I can upload the two
sessions?

Elizabeth



Hi Elizabeth,
       I'm not really sure what is going on.  It almost appears as
though
the mask you have created happens to grab values from the original
image
that are all the same value of 33195.  Your window/level screenshot is
a
little perplexing to me as I am used to seeing a histogram plot at the
top of the window/level tool panel and didn't see that in your
screenshot.  At this point there are a few things you can try, but
these
are guesses on my part:

1.  Try downloading the latest Seg3D 1.13 (I made a bad attempt at
releasing 1.13 on April 16, that resided on the web for a few days
until
being resolved on April 22) and see if that helps at all.

2.  Try saving as some other file format from ImageJ or see what
options
are given when saving from ImageJ, maybe this is doing something
unexpected to your data.

3.  Take the same set of requested screenshots and get the unmasked,
segmentation and masked data volume information when working with a
.img
file that is successful, maybe we can see a difference between a data
set that works for you and one that doesn't.  Preferably, you could do
this for the same data set using both the .img and the .tiff files.

-Kris

On 6/16/10 2:53 PM, eschn1@umbc.edu wrote:

Kris,

The info for the mask data volume is:

This is a data volume.
    Origin: 0 0 0
    Spacing: 1 1 1
    Size: 148 120 120
    Space: Unknown
    Volume: 2.1312e+006
    Min/Max: 0 33195

I tried to ftp the session to you but the computer I work at won't
allow
it, so I included the screen shot you asked for. Also, I adjusted the
window/level was to get it to show some of the different gray values.
Another thing to add about the scan I was working with: The CT scan
was
originally a .img file but I needed to reorient it in ImageJ and then
saved it as a series of .tiff files( .img wasn't a save option). Seg3D
is
only doing this to files that I first altered in ImageJ and saved as a
series of .tiff files.

Thanks again for your help.

Elizabeth








Hi Elizabeth,
       One other thing I forgot to ask for, can you also send the info
for
the masked data volume?  As far as I can tell from what you've sent
so
far everything looks like it should be working.  My current guess is
that there are a very few discrete values in the masked data so that
Seg3D applies a color map coloring these few discrete values the same
and having the data appear to be binary.  Would you mind saving your
Seg3D session and ftp'ing to ftp.sci.utah.edu ?  From there, please
put
the session file in upload/ForKris .  If this is not possible, please
select your masked data volume, open Tools->Window/Level Tool, and
take
a screenshot of the full Seg3D window.  I'm hoping the histogram
displayed by Window/Level Tool may give us some clues.  You may also
try
adjusting the window/level yourself to see whether the masked data
really is binary or is just being displayed that way.

Thanks!
Kris

On 6/16/10 11:38 AM, eschn1@umbc.edu wrote:


Hi Kris,

Thanks for your response. I use the exact same process that you use
for
Mask Data Filter so I don't have any different or additional steps
to
add.

The info for my label volume says:

This is a label volume.
Origin: 0 0 0
Spacing: 1 1 1
Size: 148 120 120
Space: Unknown
Volume: 2.1312e+006
Label Count: 45284
Label Volume: 45284
Label Volume Percentage: 2.12481

The info for my data volume says:

This is a data volume.
Origin: 0 0 0
Spacing: 1 1 1
Size: 148 120 120
Space: Unknown
Volume: 2.1312e+006
Min/Max: 30451 37939

I have also included some screen shots. Let me know if you need
anymore
information and thanks again for the help.

Elizabeth Schneider





Hello,
         I will need some more information to help troubleshoot this
issue.
Would you please click on the "i" button on the right-hand Volumes
panel
for both the label volume and data volume you are using and email
the
contents of the Volume Information that is displayed?  I am
specifically
interested in the type of volume (it should say either "This is a
data
volume." or "This is a label volume.") and the Min/Max values.

         Here is the process I use for Mask Data Filter:
1. Create segmentation of data (will be stored in a label volume)
2. Open Mask Data Filter
3. Select the label volume containing the segmented data
4. Either click the Set Mask Label button from the Mask Data panel
OR
go
to the Edit menu ->     Set Mask Label
5. Select the data volume containing the data to be masked

At this point, you should have the label volume highlighted in a
salmon/pinkish color and the data volume highlighted in a purple
color.

6.  Click the Start button on the Mask Data panel

Does this match what you are doing?  If not, please also email the
exact
steps you perform related to using the Mask Data Filter.
It also may be helpful if you can email four screenshots:  one
showing
the segmented data, one showing the data that you want to mask, one
showing the entire Seg3D window just before you click the Start
button
on the Mask Data panel, and one showing the result of the mask
data.

-Kris

On 6/14/10 9:27 AM, eschn1@umbc.edu wrote:



Hello,

I am having a problem with the Mask Data Filter. I am currently
using
Seg3D
version 1.13.0 and am still new to the program. My task is to
segment
pig
hearts from CT scans and create a gray scale mask of the pig
heart.
I
have been
using the Mask Data filter to create my final mask of the heart
but
the
weirdest thing keeps happening. The tutorial I have says that Mask
Data
is
supposed to create a gray scale mask but it will alternate between
giving me a
gray scale mask and a binary mask. I don't know what cause it to
switch
back
and forth. Right now all it's producing is binary masks when I use
it.

Does anyone know how to get Mask Data to only create a gray scale
mask
(and no
binary masks) ?
















Archive powered by MHonArc 2.6.16.

Top of page