The influence of image enhancement and reconstruction on quantitative coronary arteriography

In the coming years, cinefilm will gradually be replaced by some digital medium for the archiving of angiographic images. However, not only the question which digital archiving medium will be used in the future is important, but also which images are to be stored. Options are to either archive the raw, unprocessed images, or the enhanced images as they are displayed on the viewing monitor in the catheterization laboratory. In the first case, an off-line workstation will need additional hardware to display the images with the same image quality as they were acquired; in the second case, the question remains whether quantitative analysis programs still provide reliable results. Goal of this study was to investigate the possible effects of image enhancement and reconstruction on the results from quantitative coronary arteriographic (QCA) measurements with the Philips ACA-package (Automated Coronary Analysis). Image enhancement was achieved by an unsharp masking approach; the reconstruction of the original image from the enhanced image was attempted by an iterative deconvolution approach. The evaluation study consisted of two parts; a technical evaluation on eleven phantom tubes with known dimensions, and a clinical evaluation study on 48 coronary lesions. The results of the technical evaluation demonstrate that the measurement errors increase for the smaller vessel sizes (< 1.2 mm) when QCA is applied to reconstructed images. The systematic difference on the smallest phantom tube (0.687 mm) on unprocessed images was limited to 0.050 mm, while it increased to 0.089 mm for the reconstructed images. Moreover, the random differences for the smaller vessel sizes increased for all processed images: for 0.159 mm for the unprocessed image to 0.189 mm for the enhanced and 0.204 mm for the reconstructed image (p < 0.01). For the larger vessels, in general, no significant differences could be observed between the results of the unprocessed and processed images. The results of the clinical evaluation study demonstrate that especially the obstruction diameter is overestimated when QCA is applied to reconstructed images (0.113 mm). Although the measurements on the enhanced images did not show a significant overestimation of the obstruction diameter, the intra-observer random difference was much higher (0.199 mm for the enhanced images versus 0.140 mm for the unprocessed images, p < 0.01). In more general terms, applying QCA on enhanced images increases the random difference values, while reconstructing the original image from the enhanced images increases the systematic errors in the measured diameters. This study has clearly demonstrated that especially the smaller diameter values (< 1.2 mm) are influenced by image enhancement. Therefore, to obtain quantitative results with the desired small values for systematic and random differences, requires that the raw, unprocessed image data be archived.