Input Image Masking

Some tools in Jim (Algebra, Image Fitter, Perfusion, DCE-MRI and Dynamic Analysis tools) allow the input images to be masked so that the calculations involved are not performed for image pixels outside the mask. This saves computation time, and can make the output images more visually pleasing.

Of course, you can always mask output images after they have been created using the image Masker, but this wastes computation time and is normally less convenient, since often multiple output images need to be separately masked.

There are up to four ways to exclude (mask out) pixels, depending on the tool you are using:

These are:

to exclude pixels based on their intensity, enter an intensity threshold. If all the pixels at one position in the input images are above the threshold, analysis is performed for that pixel as normal.

use the Brain Finder exclude pixels outside the brain. If you are analysing human brain images, it may be possible to use the Brain Finder tool to first isolate the brain, and then fit to pixels only inside the brain. The Brain Finder will be applied to the first image in your series with different independent variable values, so the success of this depends on whether it has a contrast suitable for use with Brain Finder. To use this, enter a "Threshold fraction" that has previously been found to be suitable for your image (by experimenting with the Brain Finder).

to use another image to exclude pixels from analysis. An image mask in an image with the same number of columns, rows and slices of pixels as the images being analysed. For pixels to be analysed, the image mask's pixel intensity is non-zero, and for pixels that are to be excluded from analysis, the image mask's pixel intensity is zero.

to use a set of regions-of-interest (ROIs) exclude pixels from analysis. The ROIs are defined on one of the input images. Pixels outside the ROIs are not analysed.

The way in which you create the mask ROI file depends somewhat on they way your images to be analysed are organised:

• If the slice locations are contiguous in the input image(s):
  Choose one contrast or time point in the input image(s). Outline the object to be isolated on every physical slice for that one contrast or time point.
• If the contrasts or time points are contiguous in the image(s):
• If there is a single input image. Outline the object to be isolated on a single contrast/time point for every slice location. For example, if you are using the Fitter Tool to fit an exponential T₂ decay, you could outline the object on the first echo for every slice location.
• If there are multiple input images. You will need to create an additional image using the Slice Extractor and Image Concatenator. This image will contain all the physical slices with just one image contrast or time point. For example, you would extract the first image slice from each of the input images using the Slice Extractor, and then concatenate them together using the Image Concatenator. The output image from concatenation would contain as many image slices as there are physical slice locations. You would then outline the object on every slice of this image, and use this as the mask ROI file.

Save the ROIs you have created to mask the output images to an ROI file, and then select the ROI file you have just created as the Mask ROI file.

When analysis is not performed in a particular pixel because that pixel is excluded by the masking operation, the intensity in all the output images set to zero for that pixel. This can speed up the fitting considerably. It also make the output images look prettier by excluding the random values output when fitting to noise.