Edge-based segmentation of 3-D magnetic resonance images

Abstract

The 3-D visualization of MR images of the brain requires that the data set first be segmented into brain and non-brain regions. Most existing segmentation techniques either require multiple-modality data (making high-resolution imaging impractical) or require extensive user interaction (for correction of contours). We present an automated technique for the segmentation of single-modality, 3-D magnetic resonance (MR) images. Computer graphics techniques ace employed to generate 3-D views of the brain's surface from the segmented images. The segmentation algorithm is based on the 3-D difference of Gaussians (DOG) filter. A novel method was developed for the classification of regions found by the DOG filter, as well as an automated correction procedure that detects errors from the DOG filter. Spatial information is also incorporated to help discriminate between tissues. Encouraging results were obtained with an average of less than five percent error in each image. >