Advantages and limitations of two software calipers in quantitative coronary arteriography

Summary Software calipers allowing the measurement of the distances between pairs of manually defined picture elements in digitized images may be useful tools for a rapid assessment of the morphology of coronary vessels, e.g. for choosing the appropriate balloon or stent sizes before or during cardiac intervention procedures. In this paper we have studied extensively the advantages and limitations of two manual software calipers — one developed for a PC-based cinefilm analysis workstation, the other for the Philips DCI system.Based on analyses of a perspex vessel phantom with 17 sectors of known size filled with different concentrations (50 and 100%) of the contrast agent and acquired at two kV-levels (68 and 92 kV), it was found that the cinefilm approach is characterized by a very small overall (averaged over te data from three observers) systematic overestimation of 0.03 mm, and the DCI system by a systematic underestimation of 0.07 mm; the worst case accuracy value for an individual observer on frames with 100% contrast dye concentration was 0.20 mm for cinefilm, and –0.34 mm for the DCI, respectively. The overall variabilities in the measurements (precision) were almost identical for the two approaches (overall 0.07 and 0.08 mm for the cinefilm and digital approaches, respectively, and worst case for individual observers on the 100% contrast frames, 0.16 and 0.13 mm, respectively. Inverting the images (bright or dark contrast containing structures) of the phantom at 100% contrast concentration and acquired at 62 kV had no significant effect on the results obtained with the cinefilm analysis system (overall accuracy –0.12 mm for both situations), whereas it had on the results from the DCI system (overall accuracies –0.29 (dark vessels on bright background) and –0.08 mm (bright vessels on dark background), respectively). Enhancing the digital images on the DCI with unsharp masking techniques did not significantly influence the measurement accuracy and precision. Finally, it was found that woven dacron, polyurethane and polyvinylchloride catheters filled with 100% contrast dye can be measured with an overall accuracy of better than 0.13 mm on the DCI system. On the PC-based system the woven dacron and polyvinylchloride catheters would result in an overall accuracy better than 0.17 mm, and the polyurethane catheter better than 0.30 mm. The evaluation study has made clear that the nylon catheter should not be applied in QCA-studies. Thus based on the data presented and extrapolating these to the routine clinical situations, it can be concluded that the software caliper technique may be a useful tool for the rapid measurement of the size of a contrast filled structure in coronary angiograms, either from digitized cinefilm or from the digital images acquired with the Philips DCI system. However, it is to be expected that the measurements will be less accurate and precise when applied to coronary arteriograms due to additional variabilities caused by the non-uniform and rough size of coronary arterial segments. The accuracy and precision of the measurements can be improved significantly by using more sophisticated techniques with automated edge detection.