Morphological definition of anatomic shapes using minimal datasets

A new method for outlining tumour volumes and surrounding organs from computed tomography slices is presented. This task is performed during radiotherapy treatment planning and usually consists of the hand-drawing of a set of contours on the appropriate computed tomography slices of the patient lying in treatment position. This method is both time consuming and prone to errors. Our solution, based on three-dimensional scattered data interpolation, fits a three-dimensional primitive object on the contours, which are considered as 3-dimensional scattered points. The precision of this fit is iteratively refined by adding points to the interpolated contours. Interpolation using radial basis functions is combined with an appropriate mapping function between the primitive object and the scattered point cloud to provide this iterative warping. As a result, both a 3-dimensional mesh object and a set of contours are output. We describe the mapping function, as well as present some results obtained using this method