Interactive object definition in medical images using multiscale, geometric image descriptions

摘要

The medical image object definition method, which involves automatic computation of a region-based image description, along with a region containment directed acyclic graph (RCDAG) induced from the description by multiscale analysis of image structures, is described. The information resulting from this computation provides the basis for interactive object definition, during which the human user inserts semantics into the image description through additions to and alteration of the automatically computed RCDAG. A tool for interactive object definition, the Image Hierarchy Editor (IHE) is also described. Design criteria and resulting design decisions for this tool are presented, followed by a discussion of preliminary image segmentation and object definition results. For comparison papers see ibid., p.94-101 and ibid., p.102-7.