Technical Information

From NAMIC Wiki
Revision as of 14:40, 19 November 2008 by Jgumprec (talk | contribs) (→‎3D CUDA)
Jump to: navigation, search
Home < Technical Information

Technical information about the Connection of Synchrograb to Slicer

<= Back to project page

Ultrasound device

Hopefully most of this part is already done by SynchroGrab

SynchroGrab

Synchrograb was developed at Queen's University Kingston, Canada. In order create to create the 3D image it collects image data from an ultra sound device and combines them with the position data of tracking device. With the combination of images and position Synchrograb generates a 3D volume. At the end this 3D image is send to Slicer via OpenIGTLink.

In summary SynchroGrab does:

  • Continuous acquisition of ultrasound images using an open-interface ultrasound system;
  • Continuous acquisition of tracking information using a pose measurement system;
  • Explicit synchronization of the two acquisition threads;
  • Reconstruction of 3D ultrasound volumes using the synchronized data; and
  • Transfer of the raw ultrasound images, the 3D volumes and the probe tracking information to an OpenIGTLink-compliant system.

TO DO

  • Find out what is already coverd by Synchrograb
    • Adjust the tracker part, so that our tracker is working
    • Replace image acquistion part. We have to use a frame grabber card instead of reading the images directly from memory
  • Compile Synchrograb and see if everthing is working

Capture Video

  • Existing Videocard: Hauppauge ImpactVCB [1]
  • Linux Driver: Video4Linux2
    • There is existing C source for V4L2 on the web
    • Should run on new fedora kernel
Contacts
Video4Linux

Tracking device

Track data via vtk not itk

Contacts
  • Haiying Liu

Calibration

Description: To get correct information from the ultrasound device it has to be calibrated

Contacts
  • Raul San Jose

Information:

  • I have serval links and information from Raul on which I have to go through

3D CUDA

I might not use Cuda accerlation since there is no real performance advantage at this point

TO DO

  • Understand the CUDA source code of Nicholas and Ben

Contacts

Documentation: VolumeRenderingCuda Module

  • Adjust the threshold (scroll bar below 'Component Weights Scrollbars') to correct values. Especially the lower value. Normally > 0
  • Scalar Opacity Mapping: Shows a histogram of the used values an their transparence. Low transparent, high solid
  • Scalar Color Mapping: You can click into and than in the color circle above and select certain colors for a specific value
  • Composite: Here you can select different rendering modes
  • Volume: You can also select just to display a slice instead of the whole volume. You can transform the slice in the Transforms Module and than select the transformation matrix in Slice Matrix
  • Plus/Minus: Select if you want to see everything or just the volume on one side of the slice and on the other

How to "Enable CUDA 2.0 in Fedora 8 for use with Slicer 3.3 Alpha"

Date: Sept. 26th 2008

Fedora 8 is the newest Fedora version for which CUDA is available. The available CUDA Version is 2.0.
To enable CUDA 2.0 in Fedora 8 for Slicer 3.3 Alpha follow the steps below:

Installation of Drivers

  • Download the CUDA installation files (Driver, Toolkit, SDK) here [2] and remember the directory in which you downloaded them. I used

for the driver installation binary

 ~/Download/CUDA/Driver/ 

for the tool kit installation binary

 ~/Download/CUDA/Toolkit/

and for the SDK installation binary

 ~/Download/CUDA/SDK/
  • For the installation there must not run a X-Server. Therefore we have to restart the computer. When you see the GRUB Bootscreen after the reboot press any key to enter the GRUB menu. Highlight the Fedora Version you want to boot and press
 a

Then type

 Type 3

and hit the 'Enter' key. Fedora will boot up without X Server and drop you at Command Line Login prompt.

  • Login and change into root mode via:
 su
  • Then switch to the directory where you downloaded the driver. For me this means:
 cd ~/Download/CUDA/Driver/
  • Now we have to change the permissions for the driver-binary and make it executable. This works with
 chmod +a BINARY-NAME

(Replace 'BINARY-NAME' with the name of the driver binary)

  • Now start the installation with:
 ./BINARU-NAME

I changed none of the default values during the installation

  • For the sdk compilation follow this reference guide: [3]
    If you get the error message: "ld: cannot find -lglut", like I did, have a look at the reference under 'IV. Known Issues'
  • In the next step we do the same for the toolkit and for the sdk, e.g chance to the according directory, change the permissions and run the installation with the default values

Enable CUDA in Slicer 3.3 Alpha

  • Start with the installation manual of Ben Grauer. You find it here [4]
    Below you find corrections to certains points that did not work for me the way it was explained in the manual
  • In 1.1.2 Check out the following slicer branch instead and build it:
 svn co http://svn.slicer.org/Slicer3/trunk Slicer3
 ./Slicer3/Scripts/getbuildtest.tcl
  • Before you can go on with step 1.2.2 you have to Uncomment:
    • CudaSupport in Slicer3/Libs/CMakeLists.txt
    • VolumeRenderingCuda in Slicer3/Modules/CMakeLists.txt
  • In step 1.2.2
    • you have to use your build directory (Slicer-build) as parameter for ccmake
    • for CUDA_INSTALL_PREFIX use /usr/local/cuda
    • for FOUND_CUT use $HOME/NVIDIA_CUDA_SDK/common/lib/linux/libcudpp.a
    • for FOUND_CUT_INCLUDE use $HOME/NVIDIA_CUDA_SDK/common/inc
  • Step 1.2.3 was not necessary for me
  • Just do a rebuild (run Slicer3/Scripts/getbuildtest.tcl) and the VolumeRenderingCUDA module is available

This installation worked for me on a DELL machine with XEON Dual Core and a Nvidia G8800. For fast rendering make sure not enable compiz(Desktop Effects).