Contents

  1. Q: What is the format of the fiber data file?
  2. Q: How is each fiber stored?
  3. Q: How do you read in individual fiber coordinates outside the DTI Studio Environment?
  4. Q: There is an offset of 128 between 'file header' and 'fiber data'. Is there any data stored here?
  5. Q: Does the fiber coordinate represent its image matrix index or real spatial coordinate?
  6. Q: Is the offset 28 between two sequential fiber data? The first fiber length is 8, but the second one is much larger (~10^9)
  7. Q: Does DTI Studio only support the older versions of DICOM? I tried the latest Rel.2 Philips version and it didnt work.
  8. Q: DTI Studio was working on my computer but now I get the following error “Wglcreatecontext::Error”. It runs fine on other machines though.
  9. Q: How do you calculate the color map?
  10. Q: During tensor calculation I get an "out of memory" message. I have 56 directions, 256 x 256, 72 slices, 6 repetitions. Why is this?
  11. Q: Can DTI Studio read Brucker DICOM images?
  12. Q: What are the units of eigenvalues?
  13. Q: Can I use any program such as Matlab to get the FA and principle vector, save the two files and then use DTI Studio for fiber tracking?
  14. Q: Which is left and which is right?
  15. Q: How can we load ROI files defined by other software programs?
  16. Q: What is the coordinate system we should use to define the diffusion directions in DtiStudio?
  17. Q: On what kind of images does ‘skull stripping’ tool work? And what are the optimal parameters?
  18. Q: What are the optimal parameters of mutual information registration?
  19. Q: The reported values for tensor calculation results from DtiStudio ROI-based analysis has only 4 digits, Is there a way that I can get more integers in order to increase the sensitivity?

Q: What is the format of the fiber data file?

A: The Default fiber data file saved in DTI Studio has the following file format:

file header (offset 0)

Q: How is each fiber stored?

A: Each fiber is stored as follows:

Here is the definition of the data type:

Q: How do you read in individual fiber coordinates outside the DTI Studio Environment?

A: The following is an excerpt from the program to read fiber data:



Q: There is an offset of 128 between 'file header' and 'fiber data'. Is there any data stored here?

A: No. The header stores the image parameters and the extra space is left for any potential use in the future.

Q: Does the fiber coordinate represent its image matrix index or real spatial coordinate?

A: It is in Image matrix units.

Q: Is the offset 28 between two sequential fiber data? The first fiber length is 8, but the second one is much larger (~10^9)

A: No. All fiber data chains are packed one after another.

Q: Does DTI Studio only support the older versions of DICOM? I tried the latest Rel.2 Philips version and it didnt work.

A: This may happen for different DICOM versions. As far as we know, ther is no standard DICOM format for DWIs. Manufacturers may save the images in their own way.

Q: DTI Studio was working on my computer but now I get the following error “Wglcreatecontext::Error”. It runs fine on other machines though.

A: This is an OpenGL problem, a hardware related display driver issue. Usually these drivers are shipped with a new computer or graphics card. You may need to update your graphics card driver.

Q: How do you calculate the color map?

A: The color map is acquired from the FA map and principle eigen vector as follows:

(R,G,B) = FA * (Vx, Vy, Vz) , where FA is the image intensity of the FA map, (Vx, Vy, Vz) is the image intensity of the EigenVector 0 (principle eigen vector)

Q: During tensor calculation I get an "out of memory" message. I have 56 directions, 256 x 256, 72 slices, 6 repetitions. Why is this?

A: You may need more RAM to run it on your computer.

Q: Can DTI Studio read Brucker DICOM images?

A: Try loading the data using the available "DICOM" options. It works in most cases.

Q: What are the units of eigenvalues?

A: mm^2/s

Q: Can I use any program such as Matlab to get the FA and principle vector, save the two files and then use DTI Studio for fiber tracking?

A: Yes you can.

Q: Which is left and which is right?

A: All programs of MriStudio families: DtiStudio, DiffeoMap (previously known as Landmarker), and RoiEditor, follows the Radiological convention for image orientations. That means, in coronal and axial views, the right side of the image is the left hemisphere. In coronal, the patient is looking at you. We paid extra care to make sure this is the case for any circumstances.

Please note that many image analysis tools have been developed by Neurology groups, which follow the Neurology convention; the right is right. As a consequence, you may find some confusion in brain orientations.

For example, this manifests when you read the Analyze format. The default setting of the Analyze follows the Neurology convention. The Analyze format has two files; *.img and *.hdr. The *.img file contains raw image data, which can be read by the “Raw” format in our programs. However, if you do this, you find that the image is upside-down in the axial view. If you read the same file as the “Analyze” format, our program knows that they are reading an image in the Neurology convention and performs Neurology – Radiology conversion. Therefore, you should find the correct axial view.

Likewise, when you save your data in the Analyze format using our programs, they convert the original Radiology format to the Neurology format and save *.img and *.hdr files.

In rare situations, you may find a raw data in the Neurology convention, that appears upside-down. Because you can not read it as the Analyze format (lack of the *.hdr file), it is not possible to instruct our programs that it is in the Neurology convention. In this case, you can read it as a “Raw” file (it appears upside-down) and save it as an Analyze file (now you get *.img and *.hdr files). This new *.img file is now in the Radiology convention and you can read as a “Raw” file with the correct orientation.

Q: How can we load ROI files defined by other software programs?

A: Users can load ROI files such as fMRI activation map or manually draw ROIs to DtiStudio to perform fiber tracking.

When one wants to generate ROIs mannually, because the ROI-drawing capability of DtiStudio is not very strong, we recommend to use RoiEditor to define multiple 3D ROIs and then load them to DtiStudio for fiber tracking. Either way, they can be best done by using "Binary Mapping" format you can find in DtiStudio and RoiEditor.

To understand this format, first please perform fiber tracking by DtiStudio. It can be anything. Please just use two ROIs; one for "OR" and the second for "AND". After the tracking, click the "Save" icon in the "ROI Operation" section. Then you can find "Binary Mapping" as one of the options. Please choose this format and save it. If you open the directory where you saved the file, you should find three new files. Two files with *.dat extension are 3D binary files. If your image has 192x192x60 dimension, they should have the same dimension with pixels with "1" indicate the location of the ROI. The third file has *.map extension. This is an ASCII file. So, you can open it by a text editor. It should be straightfoward to understand this file. It also has an explanation section at the bottom. Once you understand this file, you can modify them to read and combine any 3D binary 1/0 ROI files for fiber tracking.

If you are using RoiEditor, you can create multiple 3D files, save them by "Binary Mapping", modify them to specify how you want to combine multiple ROIs ("OR", "AND", "NOT", "CUT" operations), read the file to DtiStudio, and fiber tracking is automatic. If you are using another software program, please make sure to save each ROI as 1/0 raw binary data. Then, you can simply specify the file path in the map file and read them into DtiStudio.

Q: What is the coordinate system we should use to define the diffusion directions in DtiStudio?

A:

1. All the users who are using the following conditions do not have to worry anything about gradient coordinate system (magnet frame, patient frame, imaging frame). Based on the patient position and imaging parameters (oblique angles), the gradient table is recalculated within the scanner on-the-fly to assure that gradient table x/y/z correspond to image x/y/z. When [1, 0, 0] is applied, it is along image X axis.

All GE scanners with all operating systems
Philips scanners with all operating systems except that "Gradient Overplus = on"
Siemens scanners after VB15 operating system

2. The following scanner users have to be worried about gradient coordinates. When oblique plane is used, the gradient table HAS TO BE recalculated because the gradient table is in the magnet x/y/z.

Philips scanners when "Gradient Opverplus = on"
Siemens scanners before VB15

If you are using DICOM or MOSAIC, you have to check "Rotate gradients if applicable" checkbox beneath the gradient table window.
If you are using Philips REC/PAR, please use "DTI_gradient_table _creator" at http://godzilla.kennedykrieger.org/~jfarrell/software_web.htm written by Jonathan Farrel.

Q: On what kind of images does ‘skull stripping’ tool work? And what are the optimal parameters?

A:

The skull stripping tool in ROIEditor is not suggested for use with T1-weighted images. In general, skull stripping will yield good results for b0 images with default parameter values. If the results are not optimal (eg. When using T2-weighted images), it is necessary to change parameters within the suggested ranges described below:

1. gamma - for intensity adjustment. Range: 0.5-1(original contrast)

2. W5 - the basic idea behind the method is to grow a surface from the inside to the boundary of brain, and W5 controls how much you want to grow it out; larger W5 will result in a larger final segmented region. Range:0.8-1.5

3. W3 - geometrically controls the smoothness of the surface; the larger the value, the smoother the surface. Range: 0.01-0.1

4. Pre Downsample and Segment Downsample- controls the number of times to downsample the data; a larger number results in a faster speed and coarser surface. Pre Downsample Values: 0 or 1. Segment Downsample Values: 1 or 2

5. Max Iteration - controls the maximum number of iterations at each downsampled level. Leave this parameter at a large value (eg. 1000) to get a smooth result.

6. The six ‘seed’ parameters - define the relative position of the initial surface, which is usually a box inside the brain. eg. If size of the data is 128x128x100, the XMin = 0.3 XMax = 0.7 indicate that the box bounds from 0.3*128 pixels to 0.7*128 pixels in x direction, and so on. Range: 0 – 1

Suggested values for parameters when using T2 images: W3=0.1, W5=1, gamma=1;

Q: What are the optimal parameters of mutual information registration?

A:

This implementation of mutual information registration method is based on an IDL program kindly provided by Jinsuh Kim and Godfrey Pearlson. The multidimensional Powell's method is used for optimization. The program interface provides following optimization parameters for adjusting:

Optimization Tolerance: This sets the criterion for termination of the optimization. The default is set to 0.002. Lowering the optimization tolerance will likely improve the registration result, however, the optimization will take longer time.

Initial Optimization Vector: The twelve values numbered from vector 1 to vector 12 provide the diagonal elements for an 12 by 12 diagonal matrix. The columns of this diagonal matrix are the initial sets of directions for optimization.

Q: The reported values for tensor calculation results from DtiStudio ROI-based analysis has only 4 digits, Is there a way that I can get more integers in order to increase the sensitivity?

A: In new version, the reported values will be automatically converted to scientific format (e.g. 7.43 E-4) for tiny numbers. This "4 digits" could happen for old versions. If it happened, you can try to use 1 as the b-value instead of its real value (e.g. 700). In this way, the reported values will be scaled up by its b-value (e.g. 700). In theory it works since b-value is a factor that divide the absolute number of the results.

faq (last edited 2014-06-30 12:13:40 by 128)