Fractional Anisotropy & Apparent Diffusion Coefficient S

[Zahir Momin]

Dear BrainVisa users and developers,

I have generated fractional anisotropy and diffusion coefficient maps of 55 gradient diffusion MRI data. I can see the axial slice represented map in BrainVisa. I can even choose a region of interest. Howerver, I can't see any features in BrainVisa where I can take the mean, median, mode, std deviation, and such mathematical calculations of my region of interest.
Thus, two questions:
1. Please let me know if Anatomist will allow me to choose a region of interest and find basic statistical information about that particular region.
2. If BrainVisa/Anatomist can't show/output that information, please let me know if this information can be obtained through any other software which can open .ima files.

Thank you for your time and help.

Zahir

[PeterK]

You can open ima files with basically any other software that has a "raw" import option. For ROI analysis you can try MNI Display, which is an ROI painting and analysis tool.

[Zahir Momin]

I am using a tool called ImageJ from NIH. It is supposed to be very powerful. It has RAW import. However, I have to specify the data type. And FLOAT is not one of them. I am assuming this is similar to 32 bit unsigned number for freactional anisotropy maps. I tried importing as 32 bit signed, 32 bit unsigned, 32 bit Real, but none of them read the actual pixel value of FA correctly. Reading the value correctly is important if I am going to do stats on it.

Do you know if MNI display overcomes this limitation. If MNI display does not, do you know of any other software that may be a better option. Essentially I want a software that is very strong from ROI stand point. I wil be going through lot of maps, as part of my research. That is why I am interesting in one really good ROI tool.

Zahir

[step1]

Zahir, have you tried on ImageJ to read the fractional anysotropy map using import and then 32 bit Real and "Little-Endian byte order" box checked ? It works well with my images.
I only have to multiply the resulting image by a constant 10^9 after to have non zero mean... Else the precision of computation of mean is not enough and I get 0 values.

[Jean-Francois Mangin]

I apologize for the lack of documentation about all this. Our ROI-module is improving little by little An increasing number of people are using the developement version in the lab, so we hope to release a better version this winter. For the moment, you may obtain the simple statistics you are talking about (mean, variance...) using the commandline AimsVoiStat. This may be a little bit tricky without our help, but we are in touch through the forum (even if this week our lab is closed, and a lot of us will be in summer holydays for a couple of weeks )

There should also be a way to get statistics about the bundles, but I am not sure it is OK with the released version. This winter for sure

[PeterK]

Zahir

ImageJ isn't a good volume/ROI analysis tool. I use Medx and Afni. I think the FA data are 16 bit unsigned int.

[Zahir Momin]

Dear Mr. Mangin,

I tried running the AimsVoiStat, but it is not finding its libraries. I have set the environment varibables using the .sh file that is included with brainvisa. I must admit I have never ran any command prompt commands of BrainVisa, but am eager to learn the command prompt stuff because there is so much there that the interface does not have. Please let me know if there is any special environment variables or anything that I need to set to run AimsVoiStat.

As a side note, I did find that I can save the histogram analysis of my region of intrest in a file. That has mean, std. dev, and other data. Does AimsVoiStat give the same information?

Zahir

[riviere]

Hi,

AimsVoiStat should find its libraries (otherwise there's a big problem in our binary package but in this case almost nothing would work in BrainVisa and I guess other people would have already complained). Are you sure you have correctly run the environment variables script ?
- you must use the script matching your shell: bash/sh or tcsh/csh
- just running the script as a program is not enough to actually modify your shell environment: bash/sh scripts must be run using

Code: Select all

. <script.sh>

(notice the "dot space" at the beginning) and tcsh/sch scripts must be run via the "source" command:

Code: Select all

source <script.csh>

Otherwise, you can easily know the data type of an image, either by looking at its attributes from BrainVisa magnigying glass icons if the image is in a BrainVisa database, or by manually using the command AimsFileInfo on the image file: the information you are looking for is the "data_type" attribute. The byte order used by most IO formats within our tools is the native byte order of the machine which wrote the image. You can also check if they need to be byte-swapped at read-time: if the "byte_swapping" attribute (given either by BrainVisa or AimsFileInfo) is 1 then byte_swapping is needed to load the image on the machine you are using.

And last, well, ahem, I must admit I just don't really know what AimsVoiStat exactly outputs: it's a command generally used in the PET community and I don't use it myself... just try it, you will see...

Denis