Objectives: The purpose of this study was to determine the variability of quantitative measurements from intracoronary ultrasound images, and the influence of the cardiac cycle on this variability, as a basis for in vivo applications.
Methods: Two observers analysed 30 MHz cross-sectional images from 96 in-vivo coronary arterial sites. By computer-assisted contour tracing we determined lumen area, vessel area, lesion area (vessel area minus lumen area) and percent obstruction (100% x lesion area/vessel area). Intra- and inter-observer and beat-to-beat variability, and systolic to diastolic differences were calculated by paired analysis.
Results: Consistent intra- and inter-observer differences (bias) were small (< or = 0.9%). Random variations in the two direct parameters were < or = 21.1%, but for the two derived parameters they were up to 40%. For all four parameters, random inter-observer variability was significantly greater (up to 119% for vessel area: 19.3 vs 8.8%) than intra-observer variability, but consistent variability was similar. Consistent beat-to-beat differences were small (< or = 1.4%), random variations were 8.9% to 17.5%. Random beat-to-beat variability for all four parameters was greater in diastolic than in systolic frames (up to 47.0% difference (11.9 vs 17.5%) for lesion area). Vascular dimensions were significantly greater in systole (2%). We found an error of 0.24 mm (2SD) for intra-observer variability of calculated mean arterial diameters, which is similar to the error described in angiographic studies (0.22 mm).
Conclusions: Quantitative measurements from intracoronary ultrasound images generally reproduce well. It is preferable to use directly measured parameters as opposed to derived parameters, as they are less subject to variability. Variability can be reduced by selecting systolic images.