Wahle A, Krauß U,
Oswald H, Fleck E:
CAR '97, in:
Proceedings of the 11th International Symposium and Exhibition on Computer Assisted Radiology and Surgery, Berlin, Excerpta Medica International Congress Series, Elsevier
Volume 1134, Page 978, 1997
Presentation of further
results from previous
study
(poster abstract only, refer to a full
paper as well) (Links)
Abstract: The accurate 3-D reconstruction of vascular systems from biplane angiographic projections provides the basis for the quantitative assessment of complex heart diseases, especially diffuse alterations. Compared to conventional Quantitative Coronary Angiography, which is restricted to local obstructions, the several distortion effects of x-ray systems have to be corrected more carefully. For example, remaining rotations or decenterizations of the biplane images cannot be accepted for 3-D analyses. On the other hand, the clinical routine does not allow phantom meassurements (i.e. imaging a regular grid) during patient examination. To overcome this problem, we developed a method to rectify angiographic images that considers both static and dynamic distortion patterns, but that needs only four markers on each image intensifier to calculate individual correction functions including centering and rotational alignment.
Periodically, aside of patient examination times, a regular grid is imaged on each x-ray system from some (2-4) reference angulations. For each of these reference images a third-order correction polynomial is calculated. Four semi-radiopaque markers are mounted in a centered square of 100mm edge length on every image intensifier, and adjusted according to the longitudinal axis of the imaging system. These markers are visible in the grid images as well as in every image of a patient examination. From the markers, the basic image rotation can be obtained. This rotation is used to calculate a specific correction polynomial for the actual image (i.e. the actual angulation). Since the coefficients of the correction polynomial highly correlate with the basic rotation, this method reliably eliminates geometric distortions even for intermediate angulations without the need of separate grid imaging. The method was validated by taking a series of grid images from 120 degrees left to 120 degrees right posterior oblique angulations, calculating the correction coefficients and the basic rotation for each of them. We could show a high correlation of the non-constant coefficients of the first, second and third order with the basic rotation as obtained from the markers