[Three dimensional reconstruction in intracoronary echography: a new system with automatic contour definition of...

The use of a new automated contour-detection system (CD) allows rapid quantification of the dimensions of coronary lumen and plaque in an entire three-dimensional set of intracoronary images (ICUS), obtained during an ECG-gated pull-back image acquisition. The aim of the study was to evaluate the reproducibility, feasibility and accuracy of this approach in the clinical setting. We examined 32 coronary stents implanted in 28 patients, mean age 59 +/- 16 years. The ICUS images were acquired during a motorized uniform-speed pull-back (0.5 mm/s) of the ultrasound catheter after angiographically guided stent optimization, recorded on videotape, and analyzed during a conventional two-dimensional ICUS examination. An ECG-gated pull-back was then performed and analyzed by a second operator using CD. The minimal lumen area and diameter of the stent and the lumen area of the reference segments were measured. Comparing the analysis on- and off-line, correlation coefficients (r) not less than 0.96, with a mean difference not higher than 0.3 +/- 1.5 were found. On the other side, the comparison between the two-dimensional analysis and the analysis performed on-line showed a good correlation between the two methods. The two-dimensional method revealed a systematic overestimation (delta = -1.3 +/- 2.3) of the area and diameter. The time required for the on-line three-dimensional analysis was 9 +/- 3 min, whereas the off-line analysis was performed within 35 +/- 10 min. Focal segments of stent underexpansion were seen in 4 cases by two-dimensional ICUS and in 3 more cases by three-dimensional ICUS, analyzed on- and off-line. In order to evaluate the reproducibility of the system, we examined ICUS images after an ECG-gated pull-back of 23 segments of coronary arteries in 20 patients. The intra and interobserver variability was evaluated. The mean relative difference between the area and volume measurements of intra and interobserver variability ranged from 0.14 to 1.51%, with a standard deviation not higher than +/- 7.17, a standard error of estimate not higher than +/- 4.85 and a correlation coefficient (r) not less than 0.98. Thus, the "contour detection" in ECG-gated three-dimensional ICUS is highly reproducible, suggesting the use of the system in studies of progression-regression of atherosclerosis. It is feasible for the on-line application as it can be performed rapidly and shows good agreement with measurements obtained by off-line CD.