AUTOMATIC DETECTION OF BRAIN CONTOURS IN MRI DATA SETS

A software procedure is presented for fully automated detection of bra in contours from single-echo 3-D MRI data, developed initially for sca ns with coronal orientation. The procedure detects structures in a hea d data volume in a hierarchical fashion. Automatic detection starts wi th a histogram-based thresholding step, whenever necessary preceded by an image intensity correction procedure. This step is followed by a m orphological procedure which refines the binary threshold mask images. Anatomical knowledge, essential for the discrimination between desire d and undesired structures, is implemented in this step through a sequ ence of conventional and novel morphological operations, using two- an d three-dimensional operations. A final step of the procedure performs overlap tests on candidate brain regions of interest in neighboring s lice images to propagate coherent 2-D brain masks through the third di mension. Results are presented for test runs of the procedure on 23 co ronal whole-brain data sets, and one sagittal whole-brain data set. Fi nally, the potential of the technique for generalization to other prob lems is discussed, as well as limitations of the technique.