Christian Riedel, Presenter: Nothing to Disclose

Hendrik Bolte, Abstract Co-Author: Nothing to Disclose

Juergen Biederer MD, Abstract Co-Author: Nothing to Disclose

Martin Heller MD, Abstract Co-Author: Nothing to Disclose

Multislice CT scanners allow high resolution volume imaging of the chest even along the axis of scan, the z-direction. Unfortunately, the maximum z-resolution is still approximatly two to three times smaller compared to the maximum resolution in the iso-center of the gantry plane in a 16-slice CT scanner. In helical CT scanning mode it is possible to compute multiple overlapping slices. This led to the idea to reformat a stack of highly overlapping slices into fewer and much thinner slices using a supperesolution algorithm. Thus, true isotropic volume imaging of the lungs could be realized.

Firstly, we measured the modulation transfer function of a 16-slice MSCT scanner in the gantry plane and along the z-direction, both at the iso-center of the gantry. Secondly, we scanned a chest phantom composed of a plastic shell containing inflated porcine lung by using a helical-scanning chest protocol that was optimized for highest spatial resolution. The parenchyma of the porcine lung was prepared with artificial lung nodules of known volume and the pulmonary vessels were perfused with contrast medium. We reconstructed 1 mm thick slices with an overlap of 90% and reformated them into 0.2mm thick slices using a so-called reconstruction based superresolution algorithm. Finally, we compared pulmonary nodule volumetry results and vascular tree segmentation outcomes for image data with and without superresolution postprocessing.

MTF measurements showed that the upper spatial resolution limit is 2.3 times higher in the axis of scan than along z-direction. Superresolution reformating allowed to even out this difference. Volumetry of the artificial lung nodules led to an average volume error of 28% without and 12% with superresolution postprocessing. Using the superresolution algorithm, the number of erronous definitions of outer nodule borders due to partial volume effects was reduced by 36%. Vascular tree segmentation led to an increase of 48% of detected branches when postprocessing was used.

The application of superresolution algorithms can significantly improve spatial resolution along the z-axis in helical MSCT scanning and so isotropic scanning can be achieved.

Riedel, C, Bolte, H, Biederer, J, Heller, M, Enhancement of Spatial Resolution in MSCT Imaging of the Chest Using Superresolution Algorithms.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4412776.html