Principles of image reconstruction with the SENSE algorithm
By undersampling k-space, each coil element sends a foldover image, i.e. with superimpositions depending on the location on the phase encoding axis. Because of their sensitivity profile differences, the images from the different elements do not present the same foldover artifacts. For instance, with two folded images and two different sensitivity profiles, an « unfolded» image of the full field of view can later be reconstructed.

 

This type of reconstruction is used by algorithms in the SENSE group.

Sensitivity profiles are established for each coil element whose covered spatial zone is systematically mapped. Spatial encoding in the direction of the phase encoding gradient is undersampled to save time (which is equivalent to reducing the field of view of the coil element). The intermediary images of each coil element thus present foldover artifacts. Thanks to the sensitivity profiles of the elements, the “unfolded” image can be calculated by deducing pixel value from the various intermediary images of each element. Coil element sensitivity profiles in phased array can be measured:

  • either by separate acquisition before the imaging sequence (SENSE) in the form of low resolution 3D acquisition over the entire field of view
  • or by self-calibration, measuring some of the missing lines in the center of the k-space (mSENSE).

Calibration is essential as it forms the basis for reconstructing the final image, and any noise in the data of an element profile will be reflected in the reconstructed image.