MTR and MTR Histograms

In this third example, we will compute magnetisation transfer ratio (MTR) images, and whole-brain MTR histograms. For this measurement we will use a pair of images: one without off-resonance saturation pulses, and one where these pulses are included. The first image is normally a regular proton-density-weighted image that may be acquired using a spin-echo or gradient-echo sequence. The second image is acquired in an identical way to the first, except that it includes additional off-resonance RF saturation pulses that affect the longitudinal magnetisation of protons in the larger molecules, such as cell membranes. These two images will be referred to as M0 and Ms respectively.

The analysis comes in three stages:

  • Removal of non-cerebral tissue from the M0 image.
  • Calculation of the MTR image.
  • Production of the MTR histogram.

Removal of Non-Cerebral Tissue

  1. The multi-slice M0 and Ms images need to be in two separate UNC, Analyze or NIFTI image files before analysis starts. Use the image converter to create these two images. If your pulses sequence puts the two images into one image file, use the Slice Extractor tool to separate them out into two image files. The two images must have the same number of rows, columns and slices.

  2. Load the proton-density-weighted image into Jim, and use the Brain Finder tool to isolate the brain and CSF from any surrounding cranial tissue on each slice of the image. The procedure for doing this is described in the section on finding the brain outline.

    Below is an example of a proton-density image before and after removal of extra-cerebral tissue.

    M0_before

    Before extra-cerebral tissue removal.

    M0_after

    After extra-cerebral tissue removal.

    Save the edited proton-density image to a new file.

Calculation of MTR

  1. Decide on an intensity threshold for MTR calculation. The MTR calculation will only be applied to pixels where the intensity is above the threshold. This prevents the calculation of MTR for pixels that just contain background noise. Load the image with the saturation pulses applied into Jim.

    Bring up the image statistics dialog to show the histogram of pixel intensities. Many images have a large number of pixels with zero intensity: ensure that you have the exclude_background check-box selected in the image statistics dialog, which will remove the large histogram peak at zero intensity. The image statistics window will now show the image intensity histogram:

    Ms_histo

    The peak to the left is from pixels outside the head that contain just noise. Choose an intensity threshold so that most of pixels that contain just noise are below the threshold. In the example above, you can see that a threshold of about 100 would remove the peak to the left. The choice is not critical, but do not choose too high a threshold, or pixels containing tissue of interest may be removed.

  2. Start the Algebra Tool. Insert the threshold chosen above into the Thresholdfield.

  3. Decide whether you are doing regional MTR estimation, or MTR histogram analysis.

    • For regional MTR analysis, a resolution in MTR of 0.1 percentage units is appropriate. From the Presets menu, select Magnetization Transfer Ratio (MTR) (0.1% bins). The MTR calculation will produce an image with intensities between 0 and 1000 (note the scaling factor of 1000 in the formula). 0 corresponds to 0% MTR, and 1000 corresponds to 100% MTR. Thus, changes of 1 intensity unit correspond to changes of 0.1% in MTR.
    • For MTR histogram calculation, a resolution in MTR of 1 percentage unit is appropriate. From the Presets menu, select Magnetization Transfer Ratio (MTR) (1% bins). The MTR calculation will produce an image with intensities between 0 and 100 (note the scaling factor of 100 in the formula). 0 corresponds to 0% MTR, and 100 corresponds to 100% MTR. Thus, changes of 1 intensity unit correspond to changes of 1% in MTR. This larger bin width will help to smooth the histogram.

  4. Load the edited (brain masked) proton-density image and the image with saturation pulses into the appropriate places in the Input Images panel by clicking on the openicon. Note that first image is the proton-density image (M0) and the second is the image with saturation pulses (Ms). Click on the apply_button button to start the MTR calculation. If you want to save the MTR image, select this from the option in the Algebra Frame.

    load_save

    However, you can always save the result later from the File menu of Jim's display.

  5. You can now do regional analysis on this MTR image, or you can go on to form the MTR histogram.

MTR Histogram

  1. If not already loaded, load the calculated MTR image into Jim, and bring up the Image Statistics window.

  2. It is usual to reduce the influence of CSF on the MTR histogram by providing a cut-off MTR value below which pixels are not included in the MTR histogram. The usual value is an MTR of 10%. Adjust the Min Contrast slider to a value of 10:

    contrast_min_10

    Note: this min. value applies if the bin width chosen was 1 MTR percentage unit; if you used a bin width of 0.1%, then the min. value should 100.
    Also set the Contrast Max slider to its maximum possible value, to ensure that no pixels with high MTR are excluded from the histogram:

    contrast_max_97

  3. Make sure you have a display layout showing more that one image slice, and that you do not have a selected slice. In this way, the image statistics shown will apply to the whole multi-slice data-set, not a specific slice. The Image Statistics viewer will now show the whole-brain MTR histogram.

  4. Click on the show_normalised check box if you want to produce a normalised MTR histogram. For normalised histograms, the total area under the MTR histogram sums to 1.0.

    mtr_histogram

  5. Write the histogram to file, for importing to your favourite statistical analysis package. From the Image Statistics File menu, choose Write: file_write. A File Chooser now pops up prompting you for a file name to save the histogram.