Cloud Models: Their Construction and Employment in Automatic MRI Segmentation of the Brain

A cloud is a triple consisting of a fuzzy object model, a delineation algorithm, and a criterion function for evaluating delineations. It employs recognition and delineation in a tightly coupled manner to accomplish image segmentation. It captures shape variations of a given object/object assembly to form an uncertainty region for its boundary. For any image position, delineation is executed in the uncertainty region to obtain a candidate object/object assembly, and the criterion function assigns a score to it. Image segmentation is defined by the candidate with the highest score. This work presents and compares three cloud models in automatic MR-T1 image segmentation of the cerebrum, the cerebellum, and cerebral hemispheres. These structures are connected in several parts, imposing serious segmentation challenges. The results show that the methods are fast, accurate, and can eliminate user intervention or, at least, reduce it to simple corrections. Their applications go beyond medical imaging to new vistas in various areas served by image segmentation.