Three-Dimensional Medical Image Registration Using a Patient Space CorrelationTechnique

摘要

The routine clinical use of three-dimensional data provided by modern medicalimaging procedures is often impeded by the difficulty in accurately correlating the resultant volume datasets. These data are frequently obtained at different times using the same modality, or images of the same patient are sometimes produced using more than one imaging modality. In order to analyze the similarities and differences between such images, it is necessary for the medical imaging data to be spatially aligned using a process known as image registration. This research investigated a structure-based image registration technique based upon simple, three-dimensional relationships among user identified landmarks. An image registration system was developed to allow a user to identify anatomic landmarks or external markers anywhere within the entire volume of the medical imaging dataset. A graphical, user-centered interface design minimizes landmark placement error. Landmarks identified in images of one volume dataset are mapped to corresponding landmarks from another volume to determine a registration transformation. The transformation is then applied to the viewing parameters of a suitable volume visualization tool. Examples are shown using a surface rendering system.