Segmentation of Anatomical Structures Using Volume Definition Tools

Abstract

The definition of structures and the extraction of an organ’s shape are essential parts of medical imaging applications. These might be applications like diagnostic imaging, image-guided surgery, or radiation therapy. The aim of the volume definition process is to delineate the specific shape of an organ on a digital image as accurately as possible, especially for 3D rendering, radiation therapy, and surgery planning. This can be done either through manual user interaction or by applying imaging processing techniques for the automatic detection of specific structures in the image using segmentation techniques. Segmentation is the process that separates an image into its important features (primitives) so that each of them can be addressed separately. This converts the planar pixel of the image into a distinguishable number of individual organs or tumors that can be clearly identified and manipulated. The segmentation process might involve complicated structures, and in this case usually only an expert can perform the task of identification manually on a slice-by-slice basis. Humans can perform this task using complex analysis of shape, intensity, position, texture, and proximity to surrounding structures. In this study we present a set of tools that are implemented in several computer-based medical applications. The central focus of this work is on the techniques used to improve the time and interaction needed by a user when defining one or more structures. These techniques involve interpolation methods for manual volume definition as well as methods for semi-automatic organ shape extraction. Finally, we will investigate volume segmentation aspects within the framework of radiation therapy.