Focus-of-attention, originally defined in [1,2,3], is the process of using knowledge to identify regions of interest and ``zoom'' in on them iteratively. Conversely, knowledge can be used to exclude clusters that are not of interest, such as separating normal tissues, such as CSF and gray matter, from the more important pathology tissues. Focus-of-attention relies on a hierarchy based on the cluster distribution described in Section 2.2 and must be flexible to handle the fuzziness of the various distribution possibilities.
Through focus-of-attention we can isolate a region (i.e., only work with pixels in the region) and perform further clustering. This is done since a focussed region will have less patterns to cluster and, generally, a smaller number of clusters to break the region into. Since reclustering is an iterative process, it is possible for these subclusters to require additional focusing or subdivision. We refer to clusters/patterns selected by this process as ``focus-of-attention clusters/patterns.''