X-Ray Imaging and Computed Tomography

Image Guidance for Radiation Therapy

Imaging in Radiation Therapy using Motion-Compensated Reconstruction

Figure 1: Example of linear accelerator for radiation therapy with an additional cone-beam CT for imaging

Today, radiation therapy is often combined with tomographic imaging. Such image-guided radiation therapy (IGRT) systems are comprised of the linear accelerator that provides the megavoltage beam for radiation therapy and an additional kilovoltage cone-beam CT (CBCT) that provides information about the patient or tumor position just before or even during the radiation treatment. This information may be used to adapt the patient position or the treatment plan. The gantry rotation speed of such IGRT devices is limited to be not faster than one minute for a full rotation. Consequently, respiratory motion of the patient does not only occur during treatment, but also during imaging.

Standard CBCT image reconstruction approaches, such as the Feldkamp reconstruction algorithm, simply backproject the entire projection data into the image volume ignoring the different motion states during acquisition and thereby suffer from motion artifacts and from blurring. To be able to determine and potentially track the position of a tumor subjected to motion, a 4D CBCT acquisition and reconstruction is required. The essential part of 4D CBCT is the retrospective gating which sorts all rawdata into different rawdata sets according to the patient’s respiratory motion phase or amplitude. A simple way to obtain 4D CBCT images is
to Feldkamp (PC-Feldkamp) approach, however, suffers from streak artifacts.

A more sophisticated respiratory-correlated reconstruction method, proposed by McKinnon and Bates (MKB), consists of two steps. First, a prior image is reconstructed by processing all rawdata and ignoring the different motion states. In a second step the motion artifacts of the prior image are reduced for each bin by adding a correction image that is obtained by processing only the projection data within the bin itself. While the MKB approach reduces the streak artifacts to a certain extent, it still suffers from noise.

We therefore developed a motion compensation approach that extracts additional information about the patient motion. These so-called motion vector fields (MVF) are extracted from the PC-Feldkamp images by using non-rigid registration algorithms with a decreased sensitivity on image artifacts. By processing the entire projection data according to the thus-determined MVFs, i.e. taking into account the different motion states, all data from each bin contribute to each 4D CBCT image in motion-compensated reconstruction. This yields a 100% dose usage resulting in low image noise and simultaneously reduced motion artifacts due to the incorporation of MVFs.

In addition to respiratory motion, cardiac motion also leads to motion artifacts during imaging and to motion uncertainties during treatment. A current research project attempts to compensate for cardiac as well as respiratory motion, which is called 5D motion compensation. The typical streak artifacts in gated, but non motion-compensated 4D CBCT reconstructions become even more severe when cardiac gating is additionally applied. Therefore respiratory and cardiac motion compensation is performed sequentially. In the first step, our algorithm compensates for respiratory motion followed by cardiac motion compensation in the second step. This two-step approach turns out to be very efficient and reduces streak artifacts significantly. Our preliminary results indicate that the proposed double motion-compensated 5D CBCT results in high quality 5D images with full dose usage.

Figure 2: Comparison of reconstructions of CBCT patient data: Feldkamp, PC Feldkamp, McKinnon-Bates, and motion-compensated reconstruction employing one of our newly-developed algorithms.
© dkfz.de

to top
powered by webEdition CMS