David M. Weber

Quantitative dual-energy coronary arteriography.

Subtraction techniques for digital cardiac imaging have been hampered by misregistration artifacts. The use of dual-energy imaging is being evaluated as a means for reducing these artifacts. Results reported previously indicate that the dual-energy technique may be useful for applications such as exercise ventriculography and general quantification tasks. The purpose of the current study is to investigate the use of dual-energy subtraction imaging for quantitative coronary arteriography. In vivo coronary vessel phantoms (0.2 to 7 mm2 in cross-sectional area) were used to study the potential advantages of tissue suppressed energy subtracted images over unsubtracted images for quantification of absolute vessel cross-sectional area when cardiac motion is present. Estimates of lumen cross-sectional area (N = 20) were determined using videodensitometric analysis of selected energy subtracted and unsubtracted images. Linear regression analysis of measured and actual cross-sectional area showed energy subtracted image data (slope = 1.06, intercept = 0.48 mm2, r = 0.99) to have improved accuracy (P less than .05) and precision (P less than .05) over unsubtracted image data (slope = 1.24, intercept = 1.07 mm2, r = 0.95).

Geometric quantitative coronary arteriography. A comparison of unsubtracted and dual energy-subtracted images.

The application of dual energy (DE) subtraction techniques to quantitative coronary arteriography (QCA) has the advantage of removing the tissue signal surrounding the vessel profile. We have compared the performance of two geometric QCA algorithms on DE-subtracted and -unsubtracted images to determine, for each, if DE subtraction is advantageous. The two algorithms under study were an edge detection algorithm and a Fourier analysis-based algorithm. For each algorithm, linear regression analysis was performed of measured cross-sectional area (CSA) versus actual CSA of coronary vessel phantoms. The edge detection algorithm was found to have improved precision (P less than .05) when applied to the DE-subtracted images. The Fourier analysis algorithm, however, was not effected by the DE subtraction. Among the unsubtracted image results, the Fourier measurements were more accurate (P less than .05) than the edge detection measurements. We conclude that the benefits to edge detection QCA of DE tissue subtraction outweigh the disadvantages of increased image noise and possible misregistration artifacts. However, the Fourier algorithm is relatively insensitive to tissue signal variations.