Automatic landmarking of cortical sulci

Abstract

The positions of the cerebral sulci projected onto a closed hull enclosing the brain tissue provide both a surface description of the brain shape, and meaningful anatomical landmarks. In [1] the sulcal positions were extracted using a thresholding technique, however the use of a single threshold can result in narrower sulci being ignored. We have developed an automatic method of sulcal identification which uses line strength measurement to enhance the contrast of the sulci. The line strength is then projected onto a closed surface surrounding the brain tissue, and line strength filters are run over flat projections of this surface, to give high line strength over the sulcal mouths. Points along the centres of the sulcal mouths are then chosen using non-maximal suppression. We have also experimented with a heat flow model as an alternative to projecting intensities perpendicular to the brain surface. This model allows information to be collected from deeper into the sulci, as it does not rely on the assumption that sulci are perpendicular to the brain surface. This new method is shown to be at least as good for a simple test image as the results from the thresholding technique, and comparisons between the techniques for a real image also indicate greater efficiency.