Quantifying Uncertainties for Prostate Image-Guided Radiotherapy: A 3D Organ Reconstruction and Registration Method

Abstract

The purpose of this paper is to present a method for volumetric reconstruction, registration and margin assignation applicable to both conventional CT scans and on board CT imaging. This method does not depend on the shape of the organs, the bony anatomy or the use of markers, and we apply it to prostate and bladder. 3D reconstructions are performed by means of spline surfaces and the 3D reconstructed surfaces are registered to a planning surface, using a multidimensional alignment from the Euclidean distance transform and the Levenberg-Marquardt optimization algorithm. Once the reconstructed surfaces are registered, we define a mean surface and obtain the corresponding variances from this mean surface. The method works properly and demonstrates that once translations are insulated by registration, residual uncertainties can be handled with the margin assigned for delineation variation and organ deformation