A fuzzy rule-based system for labelling the structures in 3D human brain magnetic resonance images

In biomedical research, it is desirable to label the brain structures of interest in magnetic resonance (MR) images for quantitative and qualitative analysis. However, the exact location of specific structures in the brain varies significantly from person to person. We have developed a fuzzy rule-based labelling system which incorporates the anatomic knowledge of experts concerning the spatial relationship between different brain landmarks for identifying several major sulci. The objective was to use the identified sulci to automatically locate the frontal lobe of the brain. A number of spatial attributes, including positions, orientation and length, were used in the fuzzy rules. A flattened 2D sulcal map of the brain surface was created from a 3D MR image to represent the locations of the sulci on the brain surface. Two human brain image sets were processed by our fuzzy system and the frontal lobes in both cases were correctly located. The brain images, along with the identified brain structures, were rendered by isosurface for better visualisation.