IMPACT OF DATA CLEANSING ON THE REGISTRATION ACCURACY OF 3D STEREOPHOTOGRAMMETRY AND 3D COMPUTED TOMOGRAPHY

摘要

The accurate registration of medical image data from a wide range of modalities may benefit from the cleansing of data prior to registration. This paper investigates various levels of data cleansing required to register untextured 3D surface files generated from computed tomography (CT) and stereophotogrammetry. Surface files from a stereophotogrammetry system and CT scan of a mannequin torso were acquired as 3D surface and volumetric data, respectively. High levels of data cleansing were necessary to improve the registration so that over 50percent of the data has a registration error of 1 mm or less. This paper reviews image registration with a particular focus on surface-based registration and the iterative closest point algorithm. The paper concludes that the removal of unwanted artefacts combined with a manual transformation matrix significantly improves the registration of stereolithography (STL) stereophotogrammetry and CT data sets to a level that could be used in medical applications.