Dipole Inversion

Having performed debackgrounding you are now ready for the final stage of QSM: dipole inversion. This step calculates the underlying tissue magnetic susceptibility from the local field shift (LFS) map.

Setting up dipole inversion for Quantitative Susceptibility Mapping

Jim uses a total variation method for dipole inversion. The settings for this stage are:

  • Setting image orientation for dipole inversion Set the image orientation. Jim will try to get the image orientation from the image. If it cannot, then select this option and set the (approximate) orientation from the drop-down list. The correct orientation is needed to get the dipole field direction, relative to the main B0 field direction.

    Note: if you need to set the image plane orientation manually, the resulting susceptibility map will be inaccurate if the true orientation is not pure axial, coronal or sagittal.

  • Setting whether the subject entered the scanner feet-first If the subject entered the scanner feet-first, then select this option. If the patient position is set incorrectly, the susceptibility values will be negated.

  • Setting whether apply MEDI You can include MEDI-weighting to the inversion process. MEDI adds a structural weighting matrix derived from the magnitude image to selectively guide spatial regularisation during dipole inversion. This ensures that changes in tissue susceptibility only occur at anatomical boundaries, helping to suppress streaking artifacts while preserving sharp structural edges.

  • Setting the regularisation parameter for the dipole inversion in Quantitative Susceptibility Mapping Set the regularisation parameter that controls the balance between accurate matching to the LFS data, and suppression of streaking artefacts.

    • Too large (over-regularisation): the algorithm over-smooths the susceptibility map. image. This dampens the local peak values in susceptibility, leading to an underestimation of true susceptibility.

    • Too small (under-regularisation): the algorithm fails to stabilise the missing frequencies at the magic angle, causing severe streaking artifacts and noise to ripple across the final susceptibility image.

  • Setting number of extra erosions of the mask during dipole inversion You can further erode the interior region mask using this setting.

    Extra erosion of the mask (peeling away a few additional voxels beyond what the background-removal algorithm requires) can be useful for suppressing residual artifacts at the boundaries of the mask. Eroding these edge voxels removes imperfectly filtered local field data, effectively preventing severe streaking artifacts from propagating into the deeper, target brain structures. This sacrifices a small amount of surface coverage to improve cleanliness of the final QSM map.

You may need to experiment with the above settings for best results with your data.

Now click the Button to initiate dipole inversion for Quantitative Susceptibility Mapping to button start the inversion. Inversion will take some time, since it is an iterative process involving multiple 3-D Fourier Transforms.

The output from this inversion stage is a new image with name taken from the input image (or first input image, if you multiple input images), but with a suffix appended to the name.

  • Chi_lambda_λ an image of estimated the magnetic susceptibility, with units of parts-per-million (PPM). λ is the actual regularisation parameter value used in the inversion.

Typical susceptibility map produced by dipole inversion

One slice of the susceptibility map, produced by dipole inversion