Wahle A, Krauß U,
Oswald H, Fleck E:
18th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), Amsterdam NL, IEEE Press
Volume 2, Page 654-655, 1996/97
Image preprocessing for 3-D reconstruction from biplane angiograms, especially concerning dynamic influences of sigmoidal effects and contrast problems (Continued) (Paper) (Images) (Links)
Abstract: The accuracies of systems for Quantitative Coronary Angiography, as well as for 3-D reconstruction from biplane angiograms, strongly depend on the quality of the used image data. To complete our highly accurate reconstruction system, we developed several methods for equalizing gray value contrasts to enhance vessel detection, and improved algorithms for elimination of geometric image distortions. Conventional rectification algorithms using a regular grid mostly neglect dynamic influences like the magnetic field of the earth. To avoid the necessity of determining the distortion for each possible angulation individually, we rectify by using a specific polynomial that is interpolated from (at least) two extrema. In first studies the rectification was sufficiently performed. Further research involving all devices of the four cath labs in German Heart Institute and Virchow-Klinikum is still under way to obtain a reliable base for statistics. The quality of the interpolated rectifications can be determined by comparison of the derived polynomial coefficients with those of the direct assessment by using the grid image for a specific orientation. Approximating the distortion patterns of a specific orientation by interpolation from extrema is feasible to provide sufficient accuracy for quantitative analyses and 3-D reconstructions from angiograms. Since the estimation of rectification data for each possible angulation in each frame of time is not applicable in clinical routine, we found a pragmatic solution with minimum expense.
Full Paper:
for
Acrobat Reader
- PDF file, 231 KB
Images:
Distorted Grids
of two Different Devices in two Different Angulations
- 432 columns, 432 rows, 33 KB
Automatic Image Rectification
- 512 columns, 2663 rows, 148 KB
Distortion Patterns in
three Levels
- 708 columns, 335 rows,
5 KB
Links: