2D nonlinear anisotropic diffusion filtering shows promise for helping in the detection of the intracranial boundary in clinical MR images. The filter can be used as a preprocessing step to an intensity-based segmentation technique. The filter decreases the intensity of tissues outside the brain and makes the intensities of the brain tissues more uniform. The resulting images should be simpler to segment than the originals, especially in the presence of RF inhomogeneity.
It was not obvious from the experiments performed here that diffusion filtering could also be used for correcting RF inhomogeneity. For now diffusion filtering will be considered only in the context of finding the intracranial boundary.