An interactive procedure for extracting features of the brain from magnetic resonance images: the lobes.

The human brain can be subdivided many ways depending upon the criteria used. One of the most basic approaches is to subdivide the cortical regions into functional units composed of the frontal, occipital, parietal, and temporal lobes. To date, the process of reliably extracting measurements of the human brain from MR images has been mostly dependent on subjective decisions made by operators tracing regions. We present a technique for the identification of the cortical lobes of the brain based upon an interactive extraction program. It requires an operator to identify a fixed set of neuroanatomical guides which are used along with segmentation information to map automatically in three dimensions the full extent of each lobe without further editing by the operator. The definitions of the lobes are based upon standard, fully accepted, neuroanatomical criteria which can readily be found in the stack of images obtained from each subject. We have used this technique in this report on a subset of images from both healthy control subjects and patients with clinically confirmed memory disorders in order to establish its interoperator and intraoperator reliability.